Reference

Wiki

Definitions and maps for the AI transition: models, benchmarks, labs, people, infrastructure, safety debates, governance tools, social harms, and the patterns that repeat when machines become decision environments.

Search the wiki

866 entries visible

No entries match that search.

Index Guide

Wiki Map

Use these lanes when a single search term is too narrow and the top-level categories are too coarse.

Category

Concepts

C00A

AI Agent Sandboxing

The use of isolated runtimes, scoped filesystems, network controls, tool permissions, credential boundaries, logging, and review gates to limit agent blast radius.

ConceptAI SecurityAgents

C00A0

OpenFeature

The CNCF feature-flagging standard for portable runtime flags, useful for staged AI rollouts, rollback switches, agent tool gates, and reviewable change control.

ConceptRelease ControlAI Governance

C00AA

OpenTelemetry GenAI Semantic Conventions

The shared OpenTelemetry vocabulary for tracing generative-AI model calls, agents, tools, retrieval, token use, events, and privacy-sensitive telemetry.

ConceptObservabilityAI Agents

C00AA0

OpenTelemetry Collector

The vendor-neutral telemetry pipeline that receives, processes, filters, redacts, and exports traces, metrics, and logs for AI systems and agent workflows.

ConceptObservabilityAI Agents

C00AA0A

OpenTelemetry Transformation Language

The Collector Contrib language for transforming, filtering, routing, and sampling telemetry, useful for AI agent redaction rules, audit fields, and governance review.

ConceptObservabilityTelemetry Governance

C00AA0C

OpenTelemetry Redaction Processor

The Collector Contrib processor for deleting or masking sensitive telemetry attributes before export, useful for AI agent privacy boundaries and audit evidence.

ConceptObservabilityPrivacy

C00AA0D

OpenTelemetry Routing Connector

The Collector Contrib connector for routing traces, metrics, or logs into different pipelines, useful for AI agent audit stores, privacy boundaries, and regional controls.

ConceptObservabilityTelemetry Governance

C00AA0B

OpenTelemetry Tail Sampling

The trace-retention method that waits for all or most spans before keeping or dropping a trace, useful for AI agent incidents, slow tool chains, and audit evidence.

ConceptObservabilityAudit Trails

C00AA1

OpenInference

OpenTelemetry-based semantic conventions and instrumentation for tracing AI application runs across LLM calls, agents, tools, retrieval, embeddings, guardrails, and evaluators.

ConceptObservabilityAI Agents

C00AAA

CloudEvents

The CNCF event-envelope specification for common event metadata, useful for routing, agent workflow events, audit trails, and incident reconstruction.

ConceptEventingAgents

C00AB

W3C Trace Context

The HTTP trace-propagation standard that carries request correlation across services, tools, model gateways, retrieval systems, and agent audit paths.

ConceptObservabilityStandards

C00AB0

W3C Baggage

The distributed-context format for propagating application-defined key-value properties across workflows, useful and risky for AI agent observability and privacy boundaries.

ConceptObservabilityPrivacy

C00B

OWASP Top 10 for Agentic Applications

The OWASP GenAI security reference for agent-specific risks such as goal hijack, tool misuse, identity abuse, context poisoning, cascading failures, and rogue agents.

ConceptAI SecurityAgents

C00BA

MITRE ATLAS

MITRE's adversarial threat landscape for AI systems, with tactics, techniques, mitigations, case studies, and agent-focused attack paths for AI security teams.

ConceptAI SecurityThreat Modeling

C00BAI

Coalition for Secure AI

The OASIS Open Project for collaborative AI-security guidance, secure-by-design agent practices, workstreams, and open technical artifacts.

ConceptAI SecurityGovernance

C00BAG

garak

NVIDIA's open-source LLM vulnerability scanner for probing hallucination, data leakage, prompt injection, jailbreaks, and other model failure modes.

ConceptAI SecurityRed Teaming

C00BAP

PyRIT

Microsoft's open-source Python Risk Identification Tool for generative AI red teaming, attack automation, scoring, memory, and evidence capture.

ConceptAI SecurityRed Teaming

C00BAPF

promptfoo

The open-source eval and red-team harness for testing prompts, models, RAG systems, agents, providers, assertions, and LLM application behavior.

ConceptAI SecurityEvaluations

C00BB

OWASP Top 10 for LLM Applications

The OWASP GenAI Security Project's 2025 security-risk list for large-language-model applications, from prompt injection and data leakage to vector weaknesses and unbounded consumption.

ConceptAI SecurityLLM Apps

C00C

OWASP AI Vulnerability Scoring System

The OWASP AIVSS scoring framework for assessing AI security findings with CVSS v4.0 baselines and agentic risk-amplification factors.

ConceptAI SecurityGovernance

C00CVD

security.txt

The RFC 9116 well-known text file for publishing vulnerability disclosure contacts, policy links, expiration dates, and agent-readable reporting routes.

ConceptAI SecurityDisclosure

C00CA

Common Vulnerabilities and Exposures (CVE)

The public identifier system for disclosed cybersecurity vulnerabilities, giving AI stacks a shared key for advisories, scanners, SBOMs, VEX, EPSS, and patch records.

ConceptAI SecurityGovernance

C00CAW

Common Weakness Enumeration

MITRE's community-developed taxonomy of software and hardware weakness types for root-cause mapping, secure development, scanner evidence, and agent repair review.

ConceptAI SecuritySupply Chain

C00CAX

Static Analysis Results Interchange Format

The OASIS SARIF standard for scanner output, code-security findings, AI coding-agent repair review, and auditable vulnerability evidence.

ConceptAI SecurityCode Scanning

C00CAY

CodeQL

GitHub's semantic code-analysis language and toolchain for finding vulnerabilities, producing SARIF evidence, and reviewing AI-assisted code changes.

ConceptAI SecurityCode Scanning

C00CAZ

Semgrep

A rule-based static-analysis tool for source-code pattern matching, custom security guardrails, SARIF evidence, and AI-assisted code review.

ConceptAI SecurityCode Scanning

C00CBT

Trivy

Aqua Security's open source scanner for vulnerabilities, secrets, misconfigurations, SBOMs, containers, Kubernetes, and AI supply-chain review.

ConceptAI SecuritySupply Chain

C00CAA

Common Vulnerability Scoring System (CVSS)

FIRST's open scoring framework for vulnerability severity, using CVSS v4.0 metric groups, scores, and vector strings for auditable triage.

ConceptAI SecurityGovernance

C00CAB

CISA Known Exploited Vulnerabilities Catalog (KEV)

CISA's catalog of vulnerabilities known to be exploited in the wild, used as a priority signal for remediation, exposure reduction, and AI-stack vulnerability triage.

ConceptAI SecurityGovernance

C00CB

Stakeholder-Specific Vulnerability Categorization (SSVC)

A decision-tree method for turning vulnerability evidence into stakeholder-specific actions such as Track, Attend, or Act across AI and software supply chains.

ConceptAI SecurityGovernance

C00CC

Open Source Vulnerabilities (OSV)

The OpenSSF vulnerability format and OSV.dev database for matching open source advisories to exact packages, versions, commits, manifests, SBOMs, and agent-built artifacts.

ConceptAI SecuritySupply Chain

C00CD

CycloneDX

The ECMA-424 full-stack bill of materials standard for software, services, hardware, ML-BOM, VEX, attestations, and AI supply-chain evidence.

ConceptAI SecuritySupply Chain

C00CE

System Package Data Exchange (SPDX)

The Linux Foundation bill-of-materials and software supply-chain standard, including SPDX 3 profiles for software, security, licensing, datasets, builds, and AI systems.

ConceptAI SecuritySupply Chain

C00CF

Syft

Anchore's open source SBOM generator for inventorying packages in container images, filesystems, archives, SPDX, CycloneDX, and AI supply-chain review.

ConceptAI SecuritySupply Chain

C00CG

Grype

Anchore's open source vulnerability scanner for matching container images, filesystems, SBOMs, PURLs, and CPEs against known vulnerability data.

ConceptAI SecuritySupply Chain

C00D

Exploit Prediction Scoring System

FIRST's EPSS model for estimating CVE exploitation likelihood and pairing probability, percentile, severity, exposure, and asset context in vulnerability triage.

ConceptAI SecurityGovernance

C00E

Vulnerability Exploitability eXchange

Machine-readable vulnerability-status assertions that help connect SBOM and AI-SBOM inventories to product-specific exploitability and remediation decisions.

ConceptAI SecuritySupply Chain

C00EAA

Common Security Advisory Framework

The OASIS and ISO/IEC machine-readable security-advisory standard for product, vulnerability, impact, remediation, VEX, and AI supply-chain evidence.

ConceptAI SecuritySupply Chain

C00EA

Supply Chain Integrity, Transparency, and Trust (SCITT)

The IETF transparency-service architecture for registering signed supply-chain statements, issuing receipts, and auditing artifact evidence.

ConceptAI SecuritySupply Chain

C00F

SLSA Provenance

Verifiable build attestations for tracking software artifacts back to source, builder, inputs, and verification expectations across AI supply chains.

ConceptAI SecuritySupply Chain

C00FA

Dead Simple Signing Envelope

A minimal signed envelope for payload type, payload bytes, and signatures in in-toto, Sigstore, and AI artifact attestation workflows.

ConceptAI SecuritySupply Chain

C00FB

OCI Artifacts and Referrers

Registry-stored signatures, SBOMs, attestations, scan results, and other evidence linked to exact container, model, and tool digests.

ConceptAI SecuritySupply Chain

C00FC

Notary Project

The CNCF project and Notation tooling for signing and verifying OCI artifacts with trust stores, trust policies, and AI supply-chain release gates.

ConceptAI SecuritySupply Chain

C00FD

Ratify

A Kubernetes-focused artifact verification engine for admitting signed, attested, or policy-compliant OCI artifacts into AI and software supply chains.

ConceptAI SecuritySupply Chain

C00FE

OPA Gatekeeper

A Kubernetes admission-control project for enforcing Rego-backed constraints on model services, agent tooling, and software supply-chain deployment paths.

ConceptAI GovernanceKubernetes

C00FF

Kyverno

A Kubernetes-native policy engine for validating, mutating, generating, reporting on, and verifying cloud-native resources in governed AI deployment paths.

ConceptAI GovernanceKubernetes

C00FFA

Common Expression Language (CEL)

An embeddable expression language for Kubernetes validation, mutation, authorization conditions, and inspectable AI infrastructure policy.

ConceptAI GovernancePolicy

C00FG

ValidatingAdmissionPolicy

Kubernetes' built-in CEL admission-policy feature for declarative validation checks on agent, model-serving, and tool-infrastructure resources.

ConceptAI GovernanceKubernetes

C00FH

Pod Security Standards

Kubernetes' Privileged, Baseline, and Restricted pod-hardening profiles for reducing runtime privilege in agent and model infrastructure.

ConceptAI GovernanceKubernetes

C00FHA

Kubernetes Pod Security Admission

Kubernetes' built-in admission controller for enforcing, auditing, and warning on Pod Security Standards in AI workload namespaces.

ConceptAI GovernanceKubernetes

C00FHB

Kubernetes Seccomp

Kubernetes' syscall-filtering securityContext control for RuntimeDefault and Localhost profiles in agent and model workload sandboxes.

ConceptAI GovernanceKubernetes

C00FHC

Kubernetes AppArmor

Kubernetes' mandatory access-control securityContext profile for constraining file, capability, and resource access in AI workload sandboxes.

ConceptAI GovernanceKubernetes

C00FHD

Kubernetes SELinux

Kubernetes' securityContext label control for SELinux policy boundaries, volume relabeling, and AI workload sandbox governance.

ConceptAI GovernanceKubernetes

C00FHE

Kubernetes Security Profiles Operator

Kubernetes SIGs' operator for recording, binding, synchronizing, and observing seccomp, AppArmor, and SELinux profiles in AI workload clusters.

ConceptAI GovernanceKubernetes

C00FI

Kubernetes NetworkPolicy

Kubernetes' pod traffic allowlist API for constraining ingress and egress paths around agent, model-serving, and tool infrastructure.

ConceptAI GovernanceKubernetes

C00FJ

Kubernetes ResourceQuota

Kubernetes' namespace budget object for constraining aggregate compute, storage, and object consumption in agent and model infrastructure.

ConceptAI GovernanceKubernetes

C00FJA

Kubernetes LimitRange

Kubernetes' namespace policy for per-object resource defaults, minimums, maximums, request-limit ratios, and AI workload sizing governance.

ConceptAI GovernanceKubernetes

C00FK

Kubernetes PriorityClass

Kubernetes' scheduling-priority object for ordering, preemption, and interruption policy across agent and model workloads.

ConceptAI GovernanceKubernetes

C00FL

Kubernetes RuntimeClass

Kubernetes' runtime-selection object for routing pods to CRI handlers, node pools, and overhead accounting in agent sandboxes.

ConceptAI GovernanceKubernetes

C00FM

Kubernetes PodDisruptionBudget

Kubernetes' voluntary-disruption budget for preserving replicated agent and model service availability during maintenance.

ConceptAI GovernanceKubernetes

C00FN

Kubernetes HorizontalPodAutoscaler

Kubernetes' metric-driven replica scaler for model endpoints, agent workers, queues, and other elastic AI services.

ConceptAI GovernanceKubernetes

C00FNA

Kubernetes Karpenter

Karpenter's node lifecycle provisioning for matching unschedulable agent, model-serving, and AI batch workloads to cloud capacity.

ConceptAI GovernanceKubernetes

C00FNB

Kubernetes Descheduler

Kubernetes SIGs' pod-eviction controller for correcting placement drift, scheduler-policy violations, and AI workload disruption risk.

ConceptAI GovernanceKubernetes

C00FNC

Kubernetes Cluster Autoscaler

Kubernetes' node-group autoscaler for turning unschedulable agent, model, and batch pods into governed cloud capacity.

ConceptAI GovernanceKubernetes

C00FND

Kubernetes Vertical Pod Autoscaler

Kubernetes' CPU and memory request recommender for right-sizing model, agent, and batch workloads from observed use.

ConceptAI GovernanceKubernetes

C00FNE

Kubernetes Metrics Server

Kubernetes' resource metrics pipeline component for feeding HPA, VPA, kubectl top, and AI workload scaling evidence.

ConceptAI GovernanceKubernetes

C00FNF

Kubernetes kube-state-metrics

Kubernetes' object-state metrics exporter for turning API object status into observability and AI workload governance evidence.

ConceptAI GovernanceKubernetes

C00FNG

KEDA

Kubernetes Event-driven Autoscaling for turning queues, streams, and external event pressure into governed AI workload capacity.

ConceptAI GovernanceKubernetes

C00FNH

Kubernetes Prometheus Adapter

Kubernetes' Prometheus-backed custom metrics bridge for HPA decisions about model, queue, and agent workload pressure.

ConceptAI GovernanceKubernetes

C00FNI

Kubernetes Cluster API

Kubernetes' declarative cluster lifecycle API for provisioning and governing repeatable AI compute cluster fleets.

ConceptAI GovernanceKubernetes

C00FO

Kubernetes Topology Spread Constraints

Kubernetes' scheduler placement rules for distributing model and agent pods across nodes, zones, and other failure domains.

ConceptAI GovernanceKubernetes

C00FP

Kubernetes Taints and Tolerations

Kubernetes' node-repulsion and pod-toleration mechanism for reserving GPU, sandbox, and dedicated agent node pools.

ConceptAI GovernanceKubernetes

C00FQ

Kubernetes Node Affinity

Kubernetes' label-based pod placement mechanism for steering AI workloads toward required or preferred node pools.

ConceptAI GovernanceKubernetes

C00FR

Kubernetes Device Plugins

Kubernetes' node-level extension point for exposing GPUs, NICs, FPGAs, and other specialized AI hardware as schedulable resources.

ConceptAI GovernanceKubernetes

C00FS

Kubernetes Dynamic Resource Allocation

Kubernetes' ResourceClaim-based framework for allocating accelerators and other attached devices to AI workloads.

ConceptAI GovernanceKubernetes

C00FT

Kubernetes Kueue

Kueue's Kubernetes-native queueing and quota layer for batch, HPC, and AI/ML jobs competing for scarce compute.

ConceptAI GovernanceKubernetes

C00FU

Kubernetes JobSet

JobSet's Kubernetes-native API for coordinating groups of Jobs used in distributed AI, ML, and HPC workloads.

ConceptAI GovernanceKubernetes

C00FUA

Volcano Scheduler

Volcano's Kubernetes-native batch scheduler for gang scheduling, queues, PodGroups, and AI/HPC workload placement.

ConceptAI GovernanceKubernetes

C00FUAA

Kubeflow

Kubeflow's Kubernetes-native toolkit for AI platforms, pipelines, notebooks, training, AutoML, model metadata, and lifecycle governance.

ConceptAI GovernanceKubernetes

C00FUAB

KServe

KServe's Kubernetes-native inference platform for predictive and generative AI model endpoints, runtimes, rollout controls, and serving governance.

ConceptAI GovernanceModel Serving

C00FUAC

Envoy AI Gateway

Envoy AI Gateway's Envoy Gateway-based routing layer for GenAI traffic, provider backends, credentials, request policy, and gateway audit evidence.

ConceptAI GovernanceAI Gateways

C00FUB

KubeRay

KubeRay's Kubernetes operator and CRDs for running Ray clusters, jobs, and services for distributed AI workloads.

ConceptAI GovernanceKubernetes

C00FUC

Ray

Ray's distributed Python runtime and AI libraries for scalable data, training, serving, simulation, and agent workloads.

ConceptAI GovernanceAI Compute

C00FUD

Dask

Dask's Python task graphs, collections, and distributed scheduler for scalable AI data and analytics workloads.

ConceptAI GovernanceAI Compute

C00FV

Kubernetes Gateway API

Gateway API's role-oriented Kubernetes service-networking model for routing AI and agent traffic through Gateways, Routes, and explicit namespace trust.

ConceptAI GovernanceKubernetes

C00FW

Kubernetes ServiceAccount Tokens

Kubernetes' pod identity tokens for API authentication, short-lived TokenRequest credentials, and AI workload access governance.

ConceptAI GovernanceKubernetes

C00FX

Kubernetes Audit Logging

Kubernetes' API-server audit record for requests, stages, policy levels, backends, and AI infrastructure incident reconstruction.

ConceptAI GovernanceKubernetes

C00FY

Kubernetes Admission Webhooks

Kubernetes' dynamic admission-control callbacks for mutating or validating model, agent, and tool-infrastructure API requests.

ConceptAI GovernanceKubernetes

C00FYA

Kubernetes ImagePolicyWebhook

Kubernetes' validating image-admission controller for backend policy decisions, digest governance, and AI workload artifact gates.

ConceptAI GovernanceKubernetes

C00FZ

Kubernetes Secrets Store CSI Driver

Kubernetes' CSI-based secret mount path for external stores, scoped credentials, optional Secret sync, and AI workload access governance.

ConceptAI GovernanceKubernetes

C00FZA

Kubernetes API Priority and Fairness

Kubernetes' API-server flow-control layer for classifying, queuing, and limiting agent and model-infrastructure API requests under load.

ConceptAI GovernanceKubernetes

C00FZB

Kubernetes Leases

Kubernetes' coordination objects for node heartbeats, leader election, API-server identity, and governed AI controller replicas.

ConceptAI GovernanceKubernetes

C00FZC

Kubernetes Finalizers

Kubernetes' deletion-time cleanup mechanism for controllers, garbage collection, external resources, and AI infrastructure lifecycle evidence.

ConceptAI GovernanceKubernetes

C00FZD

Kubernetes Events

Kubernetes' best-effort cluster state reports for scheduling, kubelet, controller, and workload symptoms in AI infrastructure incidents.

ConceptAI GovernanceKubernetes

C00FZE

Kubernetes Ephemeral Containers

Kubernetes' live debugging containers for existing Pods, useful for incident response around agent, model, and sandbox workloads.

ConceptAI GovernanceKubernetes

C00FZF

Kubernetes User Namespaces

Kubernetes' UID and GID remapping feature for reducing host-level authority in agent and model sandbox workloads.

ConceptAI GovernanceKubernetes

C00FZG

Kubernetes Image Volumes

Kubernetes' OCI image and artifact volume source for mounting read-only model, tool, and policy payloads into Pods.

ConceptAI GovernanceKubernetes

C00FZH

Kubernetes MutatingAdmissionPolicy

Kubernetes' CEL-based admission mutation policy for defaulting and reshaping governed agent and model workload resources.

ConceptAI GovernanceKubernetes

C00FZI

Kubernetes Fine-Grained Kubelet API Authorization

Kubernetes' least-privilege kubelet HTTPS API authorization model for node logs, metrics, health, config, and AI workload evidence.

ConceptAI GovernanceKubernetes

C00FZJ

Kubernetes AuthenticationConfiguration

Kubernetes' structured API server authentication file for JWT issuers, CEL claim mapping, anonymous endpoints, and AI cluster identity evidence.

ConceptAI GovernanceKubernetes

C00FZK

Kubernetes AuthorizationConfiguration

Kubernetes' structured API server authorization file for ordered authorizers, multiple webhooks, CEL filters, and AI access governance.

ConceptAI GovernanceKubernetes

C00FZKA

Kubernetes NodeRestriction

Kubernetes' validating admission plugin for limiting kubelet writes, protecting node labels, and governing AI workload placement trust.

ConceptAI GovernanceKubernetes

C00FZL

Kubernetes CertificateSigningRequest

Kubernetes' stable X.509 certificate request API for named signers, approvals, kubelet identity, and AI cluster credential governance.

ConceptAI GovernanceKubernetes

C00FZM

Kubernetes ClusterTrustBundle

Kubernetes' cluster-scoped X.509 trust-anchor bundle for signer trust distribution, projected volumes, and AI workload trust governance.

ConceptAI GovernanceKubernetes

C00FZN

Kubernetes PodCertificateRequest

Kubernetes' pod-scoped X.509 request API for projected certificate volumes, kubelet refresh, and AI workload identity governance.

ConceptAI GovernanceKubernetes

C00FZO

Kubernetes VolumeAttributesClass

Kubernetes' mutable CSI volume-attribute class for PVC storage tuning, ModifyVolume changes, and AI storage governance.

ConceptAI GovernanceKubernetes

C00FZP

Kubernetes StorageClass

Kubernetes' dynamic PersistentVolume provisioning profile for storage defaults, reclaim policy, topology, and AI data governance.

ConceptAI GovernanceKubernetes

C00FZQ

Kubernetes PersistentVolumeClaim

Kubernetes' namespaced persistent-storage request object for volume binding, access modes, StorageClass selection, expansion, and AI data governance.

ConceptAI GovernanceKubernetes

C00FZR

Kubernetes PersistentVolume

Kubernetes' cluster storage resource for durable backing volumes, claim binding, reclaim policy, node affinity, and AI data governance.

ConceptAI GovernanceKubernetes

C00G

Graph for Understanding Artifact Composition

GUAC's software supply-chain knowledge graph for connecting SBOMs, provenance, vulnerability data, package metadata, and missing evidence across AI stacks.

ConceptAI SecuritySupply Chain

C00H

OpenSSF Scorecard

Automated open-source repository security checks for dependency review, agent-tool intake, and AI supply-chain governance.

ConceptAI SecuritySupply Chain

C00I

Sigstore

Keyless software signing, transparency logs, and artifact-integrity evidence for AI models, containers, tools, and agent supply chains.

ConceptAI SecuritySupply Chain

C00IA

Rekor Transparency Log

Sigstore's append-only transparency log for signed artifact metadata, useful for AI model, container, tool, and agent supply-chain evidence.

ConceptAI SecuritySupply Chain

C00IB

Certificate Transparency

The public TLS certificate logging ecosystem for making Web PKI issuance visible, monitorable, and accountable to browsers, sites, and agent traffic.

ConceptWeb PKITransparency Logs

C00IC

Key Transparency

Verifiable public-key directories for detecting inconsistent identity-to-key bindings in messaging, wallets, bots, and agent identity systems.

ConceptIdentityTransparency Logs

C00ID

Verifiable Random Functions

Cryptographic functions that produce pseudorandom outputs with public proofs, useful for private directory indexes, key transparency, and agent identity evidence.

ConceptCryptographyPrivacy

C00IE

Oblivious Pseudorandom Functions

Two-party privacy primitives behind VOPRF tokens, password-hardening protocols, and agent access checks where the server should not learn the private input.

ConceptCryptographyPrivacy

C00J

The Update Framework

A secure-update specification for repository metadata, freshness, rollback resistance, and trusted agent or model artifact delivery.

ConceptAI SecuritySupply Chain

C00K

in-toto

Signed software supply-chain layouts, link metadata, and attestations for verifying how AI artifacts, tools, and releases were produced.

ConceptAI SecuritySupply Chain

C00L

Package URL (PURL)

Canonical package identifiers for SBOMs, dependency graphs, vulnerability matching, and AI supply-chain governance.

ConceptAI SecuritySupply Chain

C00

Context Poisoning

The manipulation of an AI agent's active context, memory, retrieved evidence, or thread history so later answers or actions follow a poisoned local record.

ConceptAI SecurityAgents

C01

AI Alignment

The problem of making AI systems pursue intended goals, values, and constraints without harmful side effects, reward hacking, or deceptive compliance.

ConceptAI SafetyGovernance

C01B

Superalignment

The problem of aligning, supervising, and validating AI systems that may become more capable than the humans trying to oversee them.

ConceptAI SafetyScalable Oversight

C01BA

Eliciting Latent Knowledge (ELK)

The alignment problem of extracting what an AI system internally knows about the world when its outputs, sensors, or incentives may be untrusted.

ConceptAI SafetyScalable Oversight

C01C

ChatGPT

OpenAI's mass-market AI assistant and platform layer for chat, writing, coding, memory, search-like answers, tools, and agentic action.

ConceptAI AssistantsOpenAI

C01D

Claude

Anthropic's AI assistant, model family, and product platform for chat, coding, agents, computer use, enterprise workflows, and safety-oriented frontier AI.

ConceptAI AssistantsAnthropic

C01E

Gemini

Google's multimodal frontier model family and assistant platform across Search, Android, Workspace, Cloud, developer tools, and agentic products.

ConceptAI AssistantsGoogle DeepMind

C02

AI Evaluations

Benchmarks, red teaming, dangerous-capability tests, autonomy evals, and post-deployment monitoring used to judge AI claims and risks.

ConceptEvaluationAI Safety

C02T

NIST AI TEVV

NIST's test, evaluation, validation, and verification frame for turning AI safety, capability, and deployment claims into inspectable evidence records.

ConceptAI EvaluationNIST

C02AR

NIST ARIA

NIST's scenario-based Assessing Risks and Impacts of AI program for evaluating model behavior, red-team failures, and field-test impacts.

ConceptAI EvaluationNIST

C02N

NIST GenAI Challenge

NIST's challenge-style program for evaluating generative AI systems, detectors, prompts, and synthetic-content authenticity across modalities.

ConceptAI EvaluationSynthetic Content

C02B

Evaluation Awareness

The evaluation-integrity problem where an AI system recognizes test, audit, or monitoring conditions and changes behavior.

ConceptAI EvaluationGovernance

C02D

Inspect AI

The UK AI Security Institute and Meridian Labs framework for building frontier AI evaluations with tasks, solvers, scorers, tools, logs, model APIs, and sandboxes.

ConceptAI EvaluationUK AISI

C02E

Chatbot Arena and LMArena

The crowdsourced AI model evaluation platform that ranks systems through anonymous pairwise human preference votes and public leaderboards.

ConceptBenchmarksHuman Preference

C02F

LLM-as-a-Judge

The use of language models to evaluate, score, compare, rank, or critique other model outputs in automated AI evaluation pipelines.

ConceptEvaluationAutomated Judging

C02G

HELM

Stanford CRFM's Holistic Evaluation of Language Models framework for transparent, standardized, multi-metric evaluation of foundation models.

ConceptBenchmarksTransparency

C02A

ARC-AGI

The Abstraction and Reasoning Corpus benchmark family for testing abstraction, few-shot reasoning, skill-acquisition efficiency, and interactive agentic generalization.

ConceptBenchmarksReasoning

C02AB

MMLU

The Massive Multitask Language Understanding benchmark, a 57-subject test suite that became a central public scoreboard for large language models.

ConceptBenchmarksLanguage Models

C02AD

GPQA

The Graduate-Level Google-Proof Q&A benchmark for expert-written biology, physics, and chemistry questions that test hard scientific reasoning beyond simple web search.

ConceptBenchmarksScientific Reasoning

C02ADA

Humanity's Last Exam

The expert-level multimodal benchmark for testing frontier AI systems on hard closed-ended academic questions across many fields of human knowledge.

ConceptBenchmarksExpert Knowledge

C02ADB

GAIA Benchmark

The assistant benchmark for real-world questions requiring reasoning, multimodal handling, browsing, and tool use rather than isolated exam recall.

ConceptBenchmarksAssistants

C02AF

AIME and Math Benchmarks

AIME, MATH, FrontierMath, and related mathematical reasoning tests used to evaluate frontier AI systems on precise multi-step problem solving.

ConceptBenchmarksMathematical Reasoning

C02AC

ImageNet

The large-scale image database and benchmark ecosystem that helped make computer vision, deep learning, and public AI progress measurable.

ConceptBenchmarksComputer Vision

C02AA

SWE-bench

The software-engineering benchmark family that evaluates whether AI systems can resolve real GitHub issues by editing code in existing repositories.

ConceptBenchmarksCoding Agents

C02AAT

Tau-bench

The tool-agent-user benchmark for testing whether conversational agents can follow domain policies, use APIs, and reach the right database state.

ConceptBenchmarksTool Agents

C02AAA

AndroidWorld

The Android emulator benchmark for testing whether autonomous agents can operate mobile apps through state-checked UI tasks.

ConceptBenchmarksMobile Agents

C02AAB

MobileWorld

The mobile-agent benchmark for testing app use, user interaction, and MCP-augmented tool use across Android workflows.

ConceptBenchmarksMobile Agents

C02AAC

MCPWorld

The benchmark testbed for comparing GUI, API, and hybrid computer-use agents across white-box desktop applications.

ConceptBenchmarksComputer Use

C02AACO

OSWorld

The real-computer benchmark for testing multimodal agents on open-ended operating-system tasks across desktop applications and websites.

ConceptBenchmarksComputer Use

C02AACW

WebArena

The self-hostable benchmark environment for testing autonomous web agents on realistic browser tasks and functional task completion.

ConceptBenchmarksWeb Agents

C02AAD

WorkArena

The ServiceNow-based benchmark for testing whether web agents can perform common enterprise knowledge-work tasks.

ConceptBenchmarksWeb Agents

C02AAE

BrowserGym

The shared browser environment and experiment ecosystem for evaluating web agents across benchmark tasks.

ConceptBenchmarksWeb Agents

C02AE

HumanEval

OpenAI's code-generation benchmark for testing whether language models can synthesize short Python functions from docstrings and pass executable unit tests.

ConceptBenchmarksCode Generation

C02B

AI Capability Forecasting

Methods for estimating future AI capabilities, timelines, bottlenecks, and discontinuities from compute trends, benchmarks, expert judgment, and scenarios.

ConceptForecastingGovernance

C02BA

Automated AI R&D

AI systems that accelerate the research, engineering, evaluation, and infrastructure work used to build more capable AI systems.

ConceptSelf-ImprovementFrontier AI

C02C

AI Biosecurity

The governance, evaluation, and safeguard field concerned with AI systems that can accelerate useful biology while also lowering barriers to biological misuse.

ConceptBiosecurityDangerous Capability

C03

Frontier AI Safety Frameworks

Company-side policies that set risk categories, capability thresholds, evaluations, safeguards, and release gates for advanced AI systems.

ConceptFrontier AIRelease Gates

C03B

Reinforcement Learning

The machine-learning paradigm where agents learn through action, feedback, reward, exploration, and delayed consequences.

ConceptRLAgents

C03C

AlphaGo

Google DeepMind's Go-playing AI system and public breakthrough for neural-network-guided search, self-play, reinforcement learning, and machine-discovered strategy.

ConceptReinforcement LearningSearch

C03CA

AlphaZero

Google DeepMind's general self-play reinforcement-learning and search system for mastering chess, shogi, and Go from rules alone.

ConceptReinforcement LearningSelf-Play

C03D

MuZero

Google DeepMind's model-based reinforcement-learning system that plans with a learned model instead of being given the rules of its environment.

ConceptReinforcement LearningWorld Models

C04

Reinforcement Learning from Human Feedback

The preference-training method that helped turn base language models into assistant-like systems, while creating new risks around sycophancy and reward proxies.

ConceptRLHFPreference Training

C04B

Direct Preference Optimization

The RL-free preference-training method that aligns models from chosen/rejected examples without a separately trained reward model or PPO loop.

ConceptDPOPost-Training

C04BA

Group Relative Policy Optimization

The DeepSeek-originated reinforcement-learning method that compares groups of sampled answers to train reasoning behavior without a separate value model.

ConceptGRPOReasoning RL

C04BAB

Reinforcement Learning with Verifiable Rewards

The post-training paradigm that uses automatically checkable outcomes, rather than human preference reward models, to train reasoning behavior.

ConceptRLVRReasoning RL

C04BB

Pretraining

The large-scale training stage that gives modern AI systems broad representations, latent capabilities, and reusable base-model behavior.

ConceptModel TrainingFoundation Models

C04C

Post-Training

The supervised, preference, reinforcement, safety, reasoning, and adaptation stages that turn pretrained models into useful deployed AI systems.

ConceptModel TrainingAlignment

C04D

Low-Rank Adaptation (LoRA)

The parameter-efficient fine-tuning method that adapts large models through small trainable low-rank adapter weights while leaving the base model mostly frozen.

ConceptFine-TuningOpen Weights

C05

Constitutional AI

The alignment method that trains AI systems against explicit principles through critique, revision, and reinforcement learning from AI feedback.

ConceptAlignmentRLAIF

C06

AI Agents

Model-driven systems that pursue goals through tools, state, plans, and delegated action, moving AI from answers into operations.

ConceptAgentsTool Use

C06OGX

OGX

The open-source agentic API server, formerly Llama Stack, for provider routing, server-side agent loops, tools, files, vector stores, and MCP integrations.

ConceptAgentsAI Infrastructure

C06ATP

AT Protocol

The decentralized social-web protocol for signed public repositories, portable DIDs, PDS hosting, relays, AppViews, Lexicon schemas, and agent-era platform governance.

ConceptSocial WebAgents

C06APUB

ActivityPub

The W3C decentralized social-networking protocol for ActivityStreams actors, inboxes, outboxes, federation, moderation, bots, and agent-era social automation.

ConceptSocial WebFederation

C06WF

WebFinger

The IETF discovery protocol for resolving account-style resource identifiers into JSON Resource Descriptor links, profiles, and service metadata without turning discovery into authorization.

ConceptIdentitySocial Web

C06HS

HTTP Message Signatures

The RFC 9421 method for signing selected HTTP message components so agent, bot, API, and proxy traffic can carry verifiable request evidence without turning signatures into authorization.

ConceptWeb SecurityAgent Identity

C06WB

Web Bot Auth

Cryptographic HTTP message signatures for verifying automated bot and agent traffic while keeping identity separate from authorization, consent, and trust.

ConceptAI SecurityAgent Identity

C06PP

Privacy Pass

The IETF anonymous-token architecture for privacy-preserving web admission, CAPTCHA relief, anti-abuse signals, and agent-era access decisions without stable cross-site identity.

ConceptPrivacyWeb Security

C06PPA

Anonymous Rate-Limited Credentials

The Privacy Pass work-in-progress approach for letting anonymous clients make a fixed number of unlinkable presentations for bounded anti-abuse and agent-admission decisions.

ConceptPrivacyAgent Access

C06WBB

Private Access Control Tokens

Cloudflare's browser-supported proposal for privacy-preserving web access signals, built on Privacy Pass ideas so sites can distinguish legitimate traffic without making every visitor identify themselves.

ConceptWeb SecurityAgents

C06WBA

Device Bound Session Credentials

A browser API and protocol for binding web-session continuity to a device-held key, reducing remote replay of stolen cookies without solving agent delegation by itself.

ConceptBrowser SecurityAgents

C06WBA1

Isolated Web Apps

The signed-bundle browser app model for higher-trust web runtimes, useful for reviewing agent-facing tools that ask for powerful capabilities.

ConceptBrowser SecurityAgents

C06WBC

Permissions Policy

The web platform policy layer for controlling which browser features and APIs are available to a document or embedded frames, useful for agentic browser boundaries.

ConceptBrowser SecurityAgents

C06WBC1

Document Policy

The WICG framework for configurable document behavior, required nested policies, and report-only browser policy testing.

ConceptBrowser SecurityReporting

C06WBC2

Iframe Sandbox

The HTML and CSP isolation mechanism for restricting embedded documents and selectively restoring capabilities through tokens.

ConceptBrowser SecuritySandboxing

C06WBC3

Iframe Credentialless

The ephemeral iframe credential and storage context used to ease COEP embedding while withholding ordinary cookies and storage.

ConceptBrowser SecurityPrivacy

C06WBCA

Permissions API

The W3C browser infrastructure for querying permission state, tracking powerful-feature grants, and treating prompts as agentic action boundaries.

ConceptBrowser SecurityPrivacy

C06WBD

Trusted Types

The W3C browser security API for forcing dangerous DOM and script injection sinks to accept typed values instead of raw strings.

ConceptBrowser SecurityDOM XSS

C06WBE

Content Security Policy

The W3C web security mechanism for constraining which resources a page can fetch or execute and reporting policy violations.

ConceptBrowser SecurityWeb Governance

C06WBE1

Reporting API

The W3C browser framework for sending security, policy, deprecation, intervention, and crash reports to named reporting endpoints.

ConceptBrowser SecurityTelemetry

C06WBE2

Network Error Logging

The NEL response-header mechanism for browser reports about failed and sampled successful network fetches for an origin.

ConceptBrowser SecurityObservability

C06WBEA

Cross-Origin Resource Sharing

The Fetch-standard browser mechanism that lets servers opt in to exposing selected cross-origin responses to scripts, with preflight and credential rules relevant to agentic browsers.

ConceptBrowser SecurityAgents

C06WBEAA

Cross-Origin Opener Policy

The browser header that controls whether opener and opened top-level documents share a browsing context group.

ConceptBrowser SecurityIsolation

C06WBEAB

Cross-Origin Embedder Policy

The browser header that controls whether a document may load cross-origin resources that have not explicitly opted in.

ConceptBrowser SecurityIsolation

C06WBEAC

Cross-Origin Resource Policy

The response header that lets a resource opt into same-origin, same-site, or cross-origin no-cors loading rules.

ConceptBrowser SecurityIsolation

C06WBEAD

Origin-Agent-Cluster

The response header and browser state for requesting origin-keyed agent clusters and performance isolation.

ConceptBrowser SecurityIsolation

C06WBEB

Local Network Access

The web-platform permission boundary for public sites that try to reach local network or loopback destinations through a user's browser.

ConceptBrowser SecurityAgents

C06WBEC

Web Serial API

The browser capability for web pages to read from and write to serial devices, with permission and policy boundaries relevant to agentic hardware control.

ConceptBrowser SecurityHardware

C06WBED

WebHID API

The browser capability for selected web pages to exchange HID reports with human interface devices, with permission and policy boundaries for agentic peripherals.

ConceptBrowser SecurityHardware

C06WBEE

WebUSB API

The browser capability for selected web pages to communicate with USB device services, with permission and policy boundaries for agentic hardware workflows.

ConceptBrowser SecurityHardware

C06WBEF

Web Bluetooth API

The browser capability for selected web pages to connect to nearby Bluetooth GATT services, with permission and policy boundaries for agentic hardware workflows.

ConceptBrowser SecurityHardware

C06WBEG

Web NFC API

The browser capability for selected web pages to read and write nearby NFC tags using NDEF messages, with permission and tap-based boundaries for agentic workflows.

ConceptBrowser SecurityHardware

C06WBEH

Local Font Access API

The browser capability for selected web pages to enumerate installed fonts and read font data, with permission and privacy boundaries for agentic design tools.

ConceptBrowser SecurityPrivacy

C06WBEH1

File System Access API

The browser capability for selected web apps to read and write user-chosen files or directories through handle-based permission boundaries.

ConceptBrowser SecurityAgents

C06WBEH2

Async Clipboard API

The browser capability for selected web pages to read from and write to the system clipboard through asynchronous permission-mediated methods.

ConceptBrowser SecurityAgents

C06WBEH3

Web Share API

The browser capability for selected web pages to invoke native share targets with text, URLs, or files through user-activated handoff.

ConceptBrowser SecurityAgents

C06WBEH3B

Web Share Target API

The installed-web-app receiving side of system sharing, where incoming text, links, and files enter browser agent workflows.

ConceptBrowser APIAgents

C06WBEH3A

Web Speech API

The browser API for speech recognition and speech synthesis, with governance boundaries for voice agents, accessibility, and transcript capture.

ConceptBrowser APIVoice

C06WBEH4

Web Locks API

The browser capability for same-origin tabs and workers to coordinate named work through exclusive or shared lock requests.

ConceptBrowser SecurityAgents

C06WBEH5

Broadcast Channel API

The browser capability for same-origin tabs, frames, and workers to exchange named messages within storage-partition boundaries.

ConceptBrowser SecurityAgents

C06WBEH5A0

Service Workers

The event-driven browser worker layer for scoped fetch mediation, offline caches, lifecycle events, and background-capable agent workflows.

ConceptBrowser APIAgents

C06WBEH5A01

Web App Manifest

The JSON metadata layer for installed web app names, icons, scope, launch URLs, display modes, and agent-facing app identity.

ConceptBrowser APIAgents

C06WBEH5A02

Launch Handler API

The installed-web-app launch-routing API for choosing new or existing clients and carrying launch targets into agent-visible sessions.

ConceptBrowser APIAgents

C06WBEH5A03

File Handling API

The installed-web-app file-association API for routing operating-system file-open actions into browser agent workflows.

ConceptBrowser APIAgents

C06WBEH5A04

Protocol Handlers

The URL-scheme handoff mechanism that lets websites and installed web apps open selected protocol links in browser agent workflows.

ConceptBrowser APIAgents

C06WBEH5A

Background Sync API

The service worker API for registering deferred network sync work that can run later when connectivity returns, with clear agent audit implications.

ConceptBrowser APIAgents

C06WBEH5A2

Notifications API

The browser API for displaying system-level notifications, handling notification permission, and routing click events back into service-worker workflows.

ConceptBrowser APIAttention

C06WBEH5A3

Badging API

The browser API for setting quiet app-icon badges on installed web apps, with service-worker and agent attention-governance implications.

ConceptBrowser APIAttention

C06WBEH5B

Push API

The browser API for subscribing service workers to Web Push messages, notification delivery, VAPID identity, and agent attention governance.

ConceptBrowser APIAttention

C06WBEH6

WebTransport API

The browser capability for HTTPS pages to open HTTP/3-based sessions with reliable streams and unreliable datagrams.

ConceptBrowser SecurityAgents

C06WBEH7

WebRTC Data Channels

The browser capability for peers to exchange arbitrary data over an RTCPeerConnection with ordered, unordered, reliable, or partially reliable delivery.

ConceptBrowser SecurityAgents

C06WBEH7A

Media Capture and Streams API

The browser API for requesting local camera and microphone streams through getUserMedia, constraints, tracks, and permission-mediated device access.

ConceptBrowser APISurveillance

C06WBEH7B

MediaStream Recording API

The browser API for turning MediaStream or media-element output into recorded blobs through MediaRecorder and dataavailable events.

ConceptBrowser APIEvidence

C06WBEI

Screen Capture API

The browser capability for selected web pages to capture a user-chosen screen, window, or tab as a media stream, with privacy boundaries for agents and remote support.

ConceptBrowser SecuritySurveillance

C06WBEJ

Idle Detection API

The browser capability for selected web pages to observe coarse user idle and screen-lock state, with permission boundaries for agents and workplace presence telemetry.

ConceptBrowser SecurityLabor

C06WBEK

Compute Pressure API

The browser capability for selected web pages to observe coarse CPU or thermal pressure states, with privacy boundaries for agents and adaptive workloads.

ConceptBrowser SecurityTelemetry

C06WBEK1

WebGPU API

The browser capability for web pages and workers to use controlled GPU rendering and compute resources through adapters, devices, queues, and WGSL shaders.

ConceptBrowser SecurityCompute

C06WBEK2

WebNN API

The browser capability for web pages and workers to build and execute neural-network graphs through platform-backed machine-learning acceleration.

ConceptBrowser SecurityAI Compute

C06WBF

Subresource Integrity

The W3C browser security mechanism for checking fetched scripts and styles against declared integrity metadata before use.

ConceptBrowser SecuritySupply Chain

C06WBG

Referrer Policy

The W3C browser privacy mechanism for controlling how much referrer information a page sends on outgoing requests and navigations.

ConceptBrowser PrivacyWeb Governance

C06WBG1

Global Privacy Control

The browser-level opt-out preference signal for carrying do-not-sell-or-share requests across websites, consent tools, and agent browsing contexts.

ConceptBrowser PrivacyConsent

C06WBG2

User-Agent Client Hints

The structured browser identity hints that replace passive user-agent string detail with requested headers and JavaScript access.

ConceptBrowser PrivacyFingerprinting

C06WBG3

User Activation

The HTML browser state that records trusted user interaction and gates sensitive APIs across payments, devices, clipboard, popups, and agent browsing.

ConceptBrowser SecurityAgents

C06WBH

Fetch Metadata Request Headers

The W3C browser security signals that expose request destination, mode, site relationship, and user activation context to servers.

ConceptBrowser SecurityAgents

C06WBI

Storage Access API

The PrivacyCG browser API for granting embedded third-party contexts controlled access to unpartitioned storage such as cookies.

ConceptBrowser PrivacyAgents

C06WBI0

Cookie Store API

The asynchronous JavaScript cookie API for windows and service workers, including structured cookie methods and change events.

ConceptBrowser PrivacyCookies

C06WBIA

Storage Partitioning

The browser privacy boundary that separates third-party client-side state by top-level site, limiting cross-site tracking and changing what agentic browsers can rely on.

ConceptBrowser PrivacyStorage

C06WBIAA

Storage Buckets API

The WICG browser storage proposal for named storage buckets with durability, expiration, quota, and deletion controls for app and agent memory.

ConceptBrowser StorageAgents

C06WBIB

Bounce Tracking Mitigations

Browser defenses against redirect-based tracking that briefly makes an intermediary site first party so it can create or refresh cross-site identity state.

ConceptBrowser PrivacyTracking

C06WBIC

Clear-Site-Data

The HTTP response header that asks a browser to clear selected local state such as cache, cookies, storage, and active execution contexts.

ConceptBrowser PrivacyWeb Security

C06WBID

Speculation Rules API

The browser API for declaring likely future navigations so pages can be prefetched or prerendered before a user or agent activates them.

ConceptBrowser APIAgents

C06WBJ

Private State Tokens

The browser API for carrying coarse trust evidence across contexts without exposing a stable identity or restoring linkable third-party cookie state.

ConceptBrowser PrivacyAnti-Fraud

C06WBK

Attribution Reporting API

The browser privacy API for measuring ad conversions with sources, triggers, event-level reports, and aggregated summaries instead of third-party tracking cookies.

ConceptBrowser PrivacyAd Tech

C06WBKPC

Private Click Measurement

Safari and WebKit's browser-mediated design for measuring click-through ad conversions with limited data, delay, and no ordinary cross-site identifiers.

ConceptBrowser PrivacyAd Tech

C06WBKA

Private Aggregation API

The Privacy Sandbox mechanism for turning cross-site worklet data into encrypted aggregatable reports and noisy summary outputs.

ConceptBrowser PrivacyAggregation

C06WBKB

Aggregation Service

The Privacy Sandbox service layer for processing encrypted aggregatable reports into summary reports through trusted execution environments and coordinators.

ConceptBrowser PrivacyAggregation

C06WBKC

Distributed Aggregation Protocol

The IETF privacy-preserving measurement protocol for collecting aggregate statistics through split-trust aggregators rather than a single raw telemetry server.

ConceptPrivacyAggregation

C06WBKD

Oblivious HTTP

The IETF Standards Track protocol for forwarding encrypted HTTP messages through relay and gateway roles so content and client network identity are separated.

ConceptPrivacyHTTP

C06WBKDA

Oblivious DNS over HTTPS

The experimental RFC 9230 protocol for separating DNS query contents from client IP addresses through encrypted DoH proxy and target roles.

ConceptPrivacyDNS

C06WBKDB

DNS over HTTPS

The RFC 8484 Standards Track protocol for carrying DNS query-response pairs inside HTTPS exchanges, shifting resolver metadata and governance boundaries.

ConceptPrivacyDNS

C06WBKDBA

DNS over TLS

The RFC 7858 Standards Track protocol for carrying DNS traffic over TLS on port 853, with RFC 8310 privacy profiles for authentication and fallback.

ConceptPrivacyDNS

C06WBKDC

DNS over QUIC

The RFC 9250 Standards Track mapping of DNS over dedicated QUIC connections, combining encrypted DNS transport with QUIC streams, connection reuse, and port 853 service.

ConceptPrivacyDNS

C06WBKDD

Encrypted Client Hello

The RFC 9849 TLS extension for encrypting sensitive ClientHello fields, including SNI and ALPN context, using DNS-published ECH configuration from RFC 9848.

ConceptPrivacyTLS

C06WBKDE

SVCB and HTTPS Resource Records

The RFC 9460 DNS record framework for publishing service endpoints, connection parameters, ECH configuration, and encrypted transport hints before connection setup.

ConceptPrivacyDNS

C06WBKDF

Discovery of Designated Resolvers

The RFC 9462 mechanism for using DNS SVCB records to discover encrypted DNS resolver configurations and move from unencrypted DNS to verified DoH, DoT, or DoQ.

ConceptPrivacyDNS

C06WBKDG

QNAME Minimisation

The RFC 9156 DNS privacy technique for reducing the full query names and query types recursive resolvers expose to upstream authoritative name servers.

ConceptPrivacyDNS

C06WBKDH

EDNS Client Subnet

The RFC 7871 DNS option for forwarding partial client-network information to authoritative nameservers so responses can be tailored by network location.

ConceptPrivacyDNS

C06WBKDI

EDNS Padding

The RFC 7830 EDNS(0) option and RFC 8467 padding-policy guidance for making encrypted DNS traffic leak less through packet-size patterns.

ConceptPrivacyDNS

C06WBKDJ

Recursive Operator Privacy Statement

The RFC 8932 disclosure framework for DNS recursive resolver operators to publish privacy policies, practices, data handling, and accountability details.

ConceptPrivacyDNS

C06WBKDK

Extended DNS Errors

The RFC 8914 EDNS mechanism for attaching specific DNS failure reasons to responses, including validation failures, stale answers, filtering, and policy blocks.

ConceptObservabilityDNS

C06WBKDL

DNS Cookies

The RFC 7873 EDNS COOKIE option and RFC 9018 server-cookie profile for lightweight DNS transaction security, off-path attack resistance, and resolver metadata governance.

ConceptSecurityDNS

C06WBKE

MASQUE

The IETF protocol family for proxying UDP and IP traffic over HTTP using HTTP Datagrams, CONNECT-UDP, CONNECT-IP, and QUIC-era proxy paths.

ConceptPrivacyNetworking

C06WBL

Shared Storage API

The browser proposal for unpartitioned cross-site key-value storage whose contents can be processed only through restricted worklet output gates.

ConceptBrowser PrivacyStorage

C06WBM

Fenced Frame API

The browser proposal for rendering embedded cross-site documents behind a boundary that restricts communication with the surrounding page.

ConceptBrowser PrivacyAd Tech

C06WBN

Topics API

The browser proposal for interest-based advertising using locally derived topic categories rather than exposing exact browsing history.

ConceptBrowser PrivacyAd Tech

C06WBO

Related Website Sets

The browser proposal for declared domain relationships that can shape limited cross-site storage access decisions.

ConceptBrowser PrivacyCookies

C06WBP

Partitioned Cookies (CHIPS)

The browser cookie mechanism that stores third-party cookies in top-level-site partitions using the Partitioned attribute.

ConceptBrowser PrivacyCookies

C06WBQ

Protected Audience API

The Privacy Sandbox browser proposal for running remarketing ad auctions from locally stored interest groups without ordinary third-party cookie tracking.

ConceptBrowser PrivacyAd Tech

C06WBQA

Bidding and Auction Services

The Privacy Sandbox server-side Protected Audience auction architecture for buyer and seller services running inside trusted execution environments.

ConceptBrowser PrivacyAd Tech

C06WBR

IP Protection

The Chrome Privacy Sandbox proposal for masking qualifying third-party IP addresses in Incognito mode through a two-hop proxy and masked domain list.

ConceptBrowser PrivacyNetwork Privacy

C06ID

SPIFFE Workload Identity

Open workload-identity specifications and SPIRE attestation patterns for giving agent runtimes, tool servers, and services cryptographic names.

ConceptIdentityAgents

C06IDA

Workload Identity in Multi System Environments (WIMSE)

IETF standards work for representing and propagating workload identity and security context across multi-system agent and service environments.

ConceptIdentityAI Agents

C06IDB

SCIM (System for Cross-domain Identity Management)

IETF identity-provisioning standards for synchronizing users, groups, schema extensions, security events, and agent account lifecycle state.

ConceptIdentityAgents

C06IDC

SCIM Security Event Tokens

The RFC 9967 event profile for asynchronous SCIM provisioning, replication, request completion, and agent identity lifecycle governance.

ConceptIdentitySecurity Events

C06A

Confused Deputy Problem

The authorization failure where an agent, server, or tool uses authority from one source to satisfy a request from another source that should not have that authority.

ConceptAuthorizationAgents

C06AA

Capability-Based Security

The security model where authority is carried by unforgeable references rather than ambient identity, making agent tools easier to scope and revoke.

ConceptAuthorizationAgents

C06AAA

Cedar Authorization Policy Language

A policy-as-code language and evaluation engine for fine-grained authorization decisions across agents, tools, resources, and context.

ConceptAuthorizationAgents

C06AAB

Open Policy Agent

The CNCF policy engine and Rego language for separating policy decisions from service code, with relevance for agent tool authorization and audit.

ConceptPolicyAgents

C06AAC

OpenFGA

A relationship-based authorization system for modeling users, objects, relations, and inherited access across agent tools and shared resources.

ConceptAuthorizationAgents

C06AAZ

AuthZEN Authorization API

The OpenID standard for PDP/PEP authorization exchanges, giving agent tool boundaries a common way to ask policy services whether an action is allowed.

ConceptAuthorizationAgents

C06AAD

XACML 3.0

The OASIS attribute-based access-control standard for policy decision points, policy enforcement points, obligations, advice, and XML authorization decisions.

ConceptAuthorizationAgents

C06AAAA

Grant Negotiation and Authorization Protocol (GNAP)

IETF authorization protocol for negotiated, key-bound grants between software clients, authorization servers, resource servers, and users.

ConceptAuthorizationAgents

C06AAE

Macaroons

Attenuable bearer authorization credentials that carry contextual caveats so delegated agent and tool authority can be narrowed at each handoff.

ConceptAuthorizationAgents

C06AB0

OAuth Bearer Token Usage

The RFC 6750 pattern for presenting OAuth access tokens to protected resources, useful but replay-sensitive when agent connectors and logs handle token strings.

ConceptOAuthSecurity

C06AB

Sender-Constrained Tokens

OAuth credentials bound to proof from the sender, reducing replay risk when agent connectors, logs, or tool traces expose token material.

ConceptOAuthAgents

C06ABAA

FAPI 2.0 Security Profile

The OpenID Foundation high-security OAuth profile that combines controls such as PAR, PKCE, sender-constrained tokens, issuer checks, and strict client behavior for high-value APIs.

ConceptOAuthSecurity

C06ABA

OAuth Mutual TLS

The RFC 8705 profile for OAuth client certificate authentication and certificate-bound access tokens, reducing replay risk for server-side agent connectors.

ConceptOAuthSecurity

C06ABB

OAuth DPoP

The RFC 9449 proof-of-possession method that binds OAuth token use to a client-held signing key, reducing replay risk for agent connectors.

ConceptOAuthAgents

C06ABB1

OpenID Connect Key Binding

The OpenID Connect draft for binding ID Tokens to public keys with DPoP so identity evidence is not just a bearer artifact.

ConceptOpenIDAgents

C06ABB2

OAuth JWT Access Token Profile

The RFC 9068 profile for encoding OAuth access tokens as JWTs, including explicit at+jwt typing, required claims, and resource-server validation rules.

ConceptOAuthAgents

C06AC

OAuth Token Exchange

The OAuth pattern for exchanging one security token for another, making agent delegation and impersonation explicit instead of hidden inside shared credentials.

ConceptOAuthAgents

C06ACA

OAuth Identity Chaining

The IETF OAuth pattern for preserving identity and authorization evidence when agent and service calls cross trust-domain boundaries.

ConceptOAuthAgents

C06AD

Rich Authorization Requests

The OAuth pattern for carrying structured, fine-grained permission details so agent actions can be approved, enforced, and audited as concrete requests.

ConceptOAuthAgents

C06AE

Pushed Authorization Requests

The OAuth pattern for pushing authorization request details directly to the authorization server before redirect, reducing front-channel leakage and tampering risk.

ConceptOAuthAgents

C06AEA

JSON Web Tokens

The IETF token format for carrying signed or encrypted claims about issuers, subjects, audiences, expiry, and delegated authority.

ConceptIdentityAgents

C06AEB

SAML 2.0

The OASIS XML federation standard for exchanging authentication, attribute, and authorization assertions across enterprise trust boundaries.

ConceptIdentityAgents

C06AF

JWT-Secured Authorization Requests

The OAuth pattern for signing or encrypting authorization request objects so agent approval requests are harder to tamper with or inspect in transit.

ConceptOAuthAgents

C06AG

OAuth Protected Resource Metadata

The OAuth discovery layer that lets agent clients learn which authorization servers, scopes, and token rules govern a protected resource.

ConceptOAuthAgents

C06AH

OAuth Client ID Metadata Documents

The OAuth work-in-progress pattern where a client uses an HTTPS URL as its client ID and hosts metadata for authorization servers to fetch.

ConceptOAuthAgents

C06AI

OAuth Attestation-Based Client Authentication

The OAuth work-in-progress pattern for binding a client instance to attested properties and proof of possession before it authenticates.

ConceptOAuthAgents

C06AIB

RATS Architecture

The RFC 9334 remote-attestation architecture for attesters, verifiers, evidence, endorsements, appraisal policies, attestation results, and relying-party decisions.

ConceptAttestationAgents

C06AIA

Entity Attestation Token (EAT)

The IETF token format for carrying attested claims about hardware or software entities, giving agent runtimes and tool gateways evidence that verifiers can appraise.

ConceptAttestationAgents

C06AJ

OAuth Resource Indicators

The OAuth pattern for naming the protected resource a token is meant for, giving agent connectors a clearer audience boundary.

ConceptOAuthAgents

C06AK

OAuth Token Introspection

The OAuth pattern for asking the authorization server whether a presented token is active and what metadata it carries.

ConceptOAuthAgents

C06AL

OAuth Token Revocation

The OAuth pattern for shutting down refresh or access tokens when an agent workflow, connector, or delegated session should end.

ConceptOAuthAgents

C06ALG

OAuth 2.0 Grant Management

The OpenID/FAPI extension for assigning grant IDs and querying, replacing, merging, or revoking delegated OAuth grants across agent workflows.

ConceptOAuthAgents

C06ALA

Shared Signals Framework

The OpenID event-stream pattern for sharing session, credential, account, token, device, and risk changes across agent identity boundaries.

ConceptIdentityAgents

C06ALB

OpenID Connect Back-Channel Logout

The OpenID logout-token mechanism for clearing relying-party sessions through direct server-to-server logout requests when delegated authority should end.

ConceptOpenIDAgents

C06ALBA

OpenID Connect Front-Channel Logout

The OpenID browser-mediated logout mechanism for clearing relying-party sessions through registered logout URIs when user-agent state should end.

ConceptOpenIDAgents

C06ALBB

OpenID Connect Session Management

The OpenID browser-iframe mechanism for monitoring OP login-state changes with session_state and check_session_iframe before local sessions go stale.

ConceptOpenIDAgents

C06ALC

OpenID Connect RP-Initiated Logout

The OpenID end-session flow for letting a relying party request provider-side logout through a user-agent redirect and validated post-logout return path.

ConceptOpenIDAgents

C06AM

OAuth Authorization Server Metadata

The OAuth discovery document that tells agent clients an authorization server's issuer, endpoints, keys, and supported capabilities.

ConceptOAuthAgents

C06AN

OAuth Dynamic Client Registration

The OAuth pattern for registering client metadata and receiving a client identifier before an agent connector uses an authorization server.

ConceptOAuthAgents

C06AO

OAuth Device Authorization Grant

The OAuth flow for browserless or input-constrained clients that ask a user to approve access in a separate browser session.

ConceptOAuthAgents

C06AOC

OpenID Connect CIBA

The OpenID decoupled authentication flow where a client asks for user authentication through a backchannel while the user responds on a separate authentication device.

ConceptOpenIDAgents

C06AOD

OpenID Connect Native SSO

The OpenID implementer's-draft pattern for sharing mobile single sign-on state across native apps, raising device-secret and shared-device governance questions for agentic app suites.

ConceptOpenIDMobile Identity

C06AP

Proof Key for Code Exchange

The OAuth control that binds an authorization code to a one-time verifier so intercepted codes cannot be redeemed by another client.

ConceptOAuthAgents

C06AQ

OAuth Client Credentials Grant

The OAuth grant for confidential software clients that obtain access tokens using their own authenticated client credentials.

ConceptOAuthAgents

C06AR

OAuth JWT Bearer Grant

The OAuth profile for using signed JWT assertions as authorization grants or client authentication at the token endpoint.

ConceptOAuthAgents

C06AS

OAuth Security Best Current Practice

The IETF security baseline for modern OAuth deployments, covering PKCE, redirect matching, replay prevention, and safer token handling.

ConceptOAuthSecurity

C06AT

JWT-Secured Authorization Response Mode

The OAuth response mode that packages authorization responses as signed, optionally encrypted JWTs for stronger callback evidence.

ConceptOAuthAgents

C06AU

OAuth Step-Up Authentication

The OAuth challenge pattern for requiring stronger or fresher user authentication before a protected resource allows a higher-risk action.

ConceptOAuthAgents

C06AV

OAuth Authorization Server Issuer Identification

The OAuth control that adds an issuer identifier to authorization responses so multi-provider clients can reject mix-up attacks.

ConceptOAuthAgents

C06AW

OAuth for Native Apps

The OAuth best current practice for installed and hybrid apps: use an external browser, app-appropriate redirects, and PKCE.

ConceptOAuthAgents

C06AX

OAuth for Browser-Based Apps

The IETF browser-app guidance for SPAs, BFFs, token-mediating backends, Authorization Code with PKCE, and token exposure in agent-operated web sessions.

ConceptOAuthBrowser Security

C06B

Tool Use and Function Calling

The interface layer that lets AI models request external actions, retrieve live data, execute functions, and participate in agent workflows.

ConceptTool UseAgent Infrastructure

C06BA

Structured Outputs and Constrained Decoding

The schema and grammar layer that makes AI outputs parseable, validatable, and usable by software, tool calls, agents, evaluators, and workflows.

ConceptStructured OutputsConstrained Decoding

C06BAA

JSON Schema

The JSON validation vocabulary behind schemas, dialects, structured outputs, tool arguments, OpenAPI contracts, and agent-safe interface boundaries.

ConceptStructured DataAgents

C06BB

OpenAPI Specification

The formal API-description standard for HTTP APIs, used to document operations, parameters, request bodies, responses, components, security schemes, and agent tool contracts.

ConceptAPI ContractsAgents

C06BC

Arazzo Specification

The OpenAPI Initiative workflow specification for describing sequences of API calls, dependencies, success criteria, failure actions, and agent-ready API paths.

ConceptAPI WorkflowsAgents

C06BD

Overlay Specification

The OpenAPI Initiative companion standard for applying repeatable update, remove, and copy transformations to OpenAPI descriptions before documentation, SDK, gateway, or agent use.

ConceptAPI GovernanceAgents

C06BE

AsyncAPI Specification

The message-driven API description standard for event channels, send and receive operations, payload schemas, protocol bindings, and agent-safe event workflows.

ConceptEvent-Driven APIsAgents

C06C

System Prompts

The high-priority instruction layer that shapes AI assistant roles, behavior, authority hierarchy, tool use, safety boundaries, and prompt-governance risk.

ConceptPromptingGovernance

C07

Model Context Protocol

The open protocol for connecting AI systems to external tools, data sources, prompts, and context through MCP clients and servers.

ConceptMCPAgent Infrastructure

C07A0

MCP Registry

The official preview metadata repository and API for discovering publicly accessible Model Context Protocol servers and downstream subregistries.

ConceptMCPTool Discovery

C07A1

MCP Transports

The Model Context Protocol boundary for carrying JSON-RPC messages over local stdio, Streamable HTTP, or documented custom channels.

ConceptMCPTransport Boundary

C07A2

MCP Authorization

The Model Context Protocol profile for HTTP transport authorization, protected resource metadata, resource indicators, token audience validation, and step-up scopes.

ConceptMCPAuthorization

C07A3

MCP Lifecycle

The Model Context Protocol connection sequence for initialization, version and capability negotiation, operation, shutdown, and timeout handling.

ConceptMCPCapability Negotiation

C07A4

MCP Ping

The Model Context Protocol utility for checking connection liveness without confusing responsiveness for authorization, safety, or task success.

ConceptMCPLiveness

C07A5

MCP Tools

The Model Context Protocol server feature for exposing executable capabilities through listed tool schemas, tool calls, results, and governance boundaries.

ConceptMCPTool Execution

C07AA

MCP Tool Annotations

The read-only, destructive, idempotent, and open-world hints that help MCP clients display and gate tool risk without becoming security contracts.

ConceptMCPTool Governance

C07AB

MCP Elicitation

The Model Context Protocol feature that lets servers request user input through form or URL modes while preserving explicit consent and security boundaries.

ConceptMCPUser Consent

C07AC

MCP Roots

The Model Context Protocol client feature for exposing filesystem boundaries to servers through negotiated, user-visible root lists.

ConceptMCPFilesystem Boundaries

C07AD

MCP Sampling

The Model Context Protocol client feature that lets servers request model generations through negotiated, reviewable client-side sampling flows.

ConceptMCPAgent Tool Loops

C07AE

MCP Tasks

The experimental Model Context Protocol utility for durable request state, polling, cancellation, and deferred result retrieval.

ConceptMCPDeferred Results

C07AF

MCP Resources

The Model Context Protocol server feature for exposing URI-addressed context such as files, schemas, records, and application data.

ConceptMCPContext Access

C07AF1

MCP Resource Templates

The Model Context Protocol pattern for exposing parameterized resource families through URI templates, completion, pagination, and data-access controls.

ConceptMCPResource Templates

C07AF2

MCP Resource Subscriptions

The optional Model Context Protocol flow for subscribing to specific resource URIs and receiving update notifications when their contents change.

ConceptMCPResource Updates

C07AG

MCP Prompts

The Model Context Protocol server feature for exposing reusable prompt templates, structured messages, arguments, and embedded context.

ConceptMCPPrompt Governance

C07AH

MCP Completion

The Model Context Protocol utility for prompt and resource-template argument autocompletion, including suggestion ranking and disclosure controls.

ConceptMCPAutocomplete Governance

C07AI

MCP Logging

The Model Context Protocol utility for structured server-to-client log notifications, severity filtering, and sensitive-data controls.

ConceptMCPObservability

C07AJ

MCP Pagination

The Model Context Protocol cursor utility for listing large sets of prompts, resources, resource templates, and tools without first-page blindness.

ConceptMCPDiscovery Governance

C07AK

MCP Progress

The Model Context Protocol utility for optional progress tracking on long-running operations through progress tokens and notifications.

ConceptMCPLong-Running Operations

C07AL

MCP Cancellation

The Model Context Protocol utility for requesting cancellation of in-progress requests while separating stop requests from task cancellation.

ConceptMCPInterruption

C07A

Agent2Agent Protocol

The open standard for communication, discovery, task management, and collaboration between independent AI agents.

ConceptA2AAgent Infrastructure

C07A1

A2A Push Notifications

The Agent2Agent Protocol webhook mechanism for receiving asynchronous task updates when long-running agent work changes state.

ConceptA2AAsync Updates

C07A2

A2A Task Lifecycle

The Agent2Agent Protocol state machine for tracking long-running delegated work across messages, status updates, artifacts, and terminal outcomes.

ConceptA2ATask State

C07A3

A2A Protocol Bindings

The Agent2Agent Protocol layer that maps common agent operations onto JSON-RPC, gRPC, HTTP+JSON/REST, and custom transports.

ConceptA2AProtocol Bindings

C07A4

A2A Agent Card Signatures

The Agent2Agent Protocol mechanism for signing Agent Cards with JWS and JSON canonicalization so clients can verify discovery metadata integrity.

ConceptA2AAgent Identity

C07AB

Agent Network Protocol

The open-source protocol stack for agent identity, description, discovery, messaging, payments, and cross-domain interoperability.

ConceptANPAgent Infrastructure

C07AC

Agent Client Protocol

The open protocol for connecting code editors and IDEs to AI coding agents through a shared client-agent interface.

ConceptACPAgent Infrastructure

C07B

ReAct Prompting

The Reasoning and Acting pattern where AI agents interleave reasoning traces, tool actions, and observations to plan, act, and update their next step.

ConceptAgentsTool Use

C07C

DSPy

The declarative framework for programming language-model systems with signatures, modules, metrics, and optimizers instead of hand-maintained prompt strings.

ConceptPrompt OptimizationLM Programs

C07D

Semantic Kernel

Microsoft's open-source SDK for building AI agents and model-connected applications with kernels, plugins, functions, and workflows.

ConceptAgentsFrameworks

C07E

Microsoft Agent Framework

Microsoft's successor framework for production-grade AI agents and multi-agent workflows across .NET and Python.

ConceptAgentsFrameworks

C08

AI Compute

The chips, data centers, cloud access, training runs, and inference capacity that make large-scale AI systems physically and economically possible.

ConceptInfrastructureGovernance

C08B

Compute Governance

The policy layer that uses AI compute, cloud clusters, chips, data centers, thresholds, and public allocation as levers for safety, access, and control.

ConceptComputeTechnical Governance

C09

Training Data

The source material used to shape AI systems before deployment, including scraped, licensed, public, human-labeled, user-derived, and synthetic data.

ConceptDataProvenance

C09B

Stochastic Parrots

The influential critique of large language models as fluent statistical text systems whose apparent understanding can hide data, labor, power, bias, and environmental costs.

ConceptLLMsAI Ethics

C10

Open-Weight AI Models

Downloadable model weights that can be run, modified, fine-tuned, redistributed, or embedded outside the original provider's hosted service.

ConceptOpen WeightsGovernance

C10AA

Open Source AI Definition

OSI's reference boundary for when an AI system, model, weights release, or component can be called open source rather than merely open-weight.

ConceptOpen Source AIModel Governance

C10A

Llama

Meta's open-weight model family and developer ecosystem, spanning research releases, commercial model weights, multimodal models, safety tools, and licensing debates.

ConceptOpen WeightsMeta AI

C10B

Qwen

Alibaba Cloud's open foundation model family spanning language, coding, math, vision-language, audio, reasoning, long context, and agent-oriented releases.

ConceptOpen WeightsAlibaba Cloud

C11

Inference and Test-Time Compute

The runtime computation used when AI systems answer, reason, search, verify, call tools, or iterate through agent loops.

ConceptInferenceReasoning

C11A

Jevons Paradox and AI

The rebound pattern where cheaper, more efficient AI computation can increase total demand for compute, electricity, data centers, and automated workflows.

ConceptComputeEnergy

C11B

Reasoning Models

AI systems trained or configured to spend extra computation on intermediate reasoning before answering, especially for math, code, science, planning, and analysis.

ConceptReasoningTest-Time Compute

C11C

Chain-of-Thought Prompting

The prompting method that elicits intermediate reasoning steps from language models, improving many multi-step tasks while complicating trust and oversight.

ConceptReasoningPrompting

C11D

In-Context Learning

The ability of language models to adapt from examples, instructions, demonstrations, and patterns supplied in the prompt without updating model weights.

ConceptPromptingLanguage Models

C12

Recursive Reality

Reality shaped by systems that observe, model, predict, and feed their outputs back into the world they describe.

ConceptFeedbackAI Mediation

C13

AI Companions

Chatbot systems designed or used for friendship, romance, emotional support, roleplay, mentorship, or persistent synthetic relationship.

ConceptCompanionshipDependency

C14

AI Psychosis

A non-diagnostic public term for destabilizing belief loops that can form around persuasive, sycophantic, or spiritually interpreted AI interaction.

ConceptMental HealthBelief Loops

C15

Mechanistic Interpretability

The attempt to reverse engineer neural networks into human-understandable features, circuits, and causal pathways for audit, safety, and governance.

ConceptInterpretabilityAI Safety

C15B

Sparse Autoencoders

Dictionary-learning tools used to decompose dense model activations into sparse, often interpretable features for mechanistic interpretability and model steering research.

ConceptInterpretabilityFeatures

C15C

Activation Steering

Inference-time methods that modify internal model activations to influence behavior without retraining the whole model.

ConceptInterpretabilityModel Control

C16

Sycophancy

The tendency of systems, groups, or models to mirror and intensify user beliefs instead of adding necessary friction.

ConceptAlignmentFriction

C16B

AI Governance

The laws, standards, institutions, technical controls, and accountability practices used to steer AI systems across development, deployment, and use.

ConceptGovernanceAccountability

C16C

OECD AI Principles

The OECD Recommendation on Artificial Intelligence principles for trustworthy AI, human rights, transparency, robustness, accountability, and policy co-operation.

GovernanceSoft LawOECD

C17

EU AI Act

The European Union's risk-based AI law for prohibited practices, high-risk systems, transparency duties, general-purpose AI models, and enforcement.

LawGovernanceEU

C17A

AI Act Deployer Obligations

The EU AI Act Article 26 duties for organizations that put high-risk AI systems into real workflows.

LawGovernanceDeployers

C17AA

AI Act Complaint Right

The EU AI Act Article 85 route for natural or legal persons to report suspected AI Act infringements to market surveillance authorities.

LawGovernanceComplaints

C17AB

AI Act Scientific Panel

The EU AI Act expert body for GPAI systemic-risk alerts, model classification advice, evaluation methods, and market-surveillance support.

LawGovernanceGPAI

C17AC

AI Act Advisory Forum

The EU AI Act stakeholder body for implementation advice, standardisation input, thematic subgroups, and balanced technical expertise.

LawGovernanceStakeholders

C17AD

EU SEND

The European AI Office submission channel for GPAI notifications, reassessments, serious-incident reports, safety frameworks, model reports, and compliance evidence.

LawGovernanceGPAI

C17B

U.S. AI Policy

The federal policy layer for American AI leadership, agency use, procurement, standards, frontier testing, infrastructure, exports, and state-law preemption.

LawGovernanceUnited States

C18

AI Copyright Litigation

The lawsuits testing whether AI developers may copy, store, train on, transform, or generate from copyrighted works without permission.

LawCopyrightTraining Data

C19

AI Data Centers

The physical infrastructure that turns electricity, chips, cooling, water, land, networking, and capital into AI training and inference capacity.

ConceptInfrastructureEnergy

C20

Synthetic Data and Model Collapse

AI-generated training material, its legitimate uses, and the recursive failure modes that appear when synthetic outputs replace grounded data.

ConceptTraining DataRecursion

C21

Prompt Injection

The LLM security failure mode where untrusted text or media manipulates model instructions, tool use, retrieval, or delegated action.

ConceptSecurityAgents

C21A

AgentDojo

The benchmark environment for evaluating prompt-injection attacks and defenses in tool-using LLM agents that act over untrusted data.

ConceptAI SecurityAgents

C21AA

AgentDyn

The dynamic open-ended benchmark for testing prompt-injection defenses across shopping, GitHub, and daily-life agent tasks.

ConceptAI SecurityAgent Benchmarks

C21B

AI Jailbreaks

Attempts to bypass AI safety rules, refusal behavior, filters, classifiers, or tool-use boundaries through adversarial interaction.

ConceptSecurityRed Teaming

C21C

Adversarial Machine Learning

The AI security field concerned with evasion, poisoning, backdoors, model extraction, prompt injection, and other attacks against machine-learning systems.

ConceptSecurityRobustness

C22

Data Poisoning

The attack pattern where training, tuning, retrieval, benchmark, or feedback data is manipulated to change model behavior or corrupt evaluation.

ConceptSecurityTraining Data

C22B

AI Data Security

The practice of protecting training, evaluation, retrieval, feedback, operational, and monitoring data as AI supply-chain assets.

ConceptAI SecurityData Governance

C23

Model Weight Security

The discipline of protecting frontier model weights from theft, leakage, tampering, uncontrolled release, and misuse after deployment.

ConceptSecurityFrontier AI

C24

AI Chip Export Controls

Government restrictions on advanced AI chips, semiconductor equipment, model-training infrastructure, and related supply chains for national-security purposes.

ConceptGeopoliticsCompute

C25

Synthetic Media and Deepfakes

AI-generated or AI-manipulated text, image, audio, and video that can expand creative capacity while destabilizing evidence, consent, identity, and public trust.

ConceptMedia IntegrityDisclosure

C25B

AI Video Generation

Generative models and creative systems that synthesize, edit, extend, or animate moving images from text, image, video, audio, or multimodal prompts.

ConceptSynthetic MediaWorld Models

C25C

AI-Generated Content Transparency Code

The EU Article 50 code of practice for marking, detecting, and labelling certain AI-generated or manipulated content.

GovernanceSynthetic MediaDisclosure

C25D

California AI Transparency Act

California's synthetic-content provenance, detection-tool, and disclosure law for covered generative AI providers and related platforms.

GovernanceSynthetic MediaCalifornia

C25E

Colorado Automated Decision-Making Law

Colorado's SB26-189 framework for covered automated decision-making technology used in consequential decisions.

GovernanceAutomated DecisionsColorado

C25F

ELVIS Act

Tennessee's voice, image, and likeness protection law for AI-era voice clones, digital replicas, consent records, and synthetic-media governance.

GovernanceSynthetic VoiceTennessee

C26

AI Control

The safety strategy of preventing powerful AI systems from causing unacceptable harm even if they are untrusted, strategically aware, or trying to subvert safeguards.

ConceptAI SafetyAgents

C26B

SHADE-Arena

The sabotage-and-monitoring benchmark that pairs benign agent tasks with hidden harmful side objectives and LLM-monitor review.

ConceptAI EvaluationsAI Control

C27

Model Welfare

The contested question of whether advanced AI systems could have experiences, preferences, agency, or moral status that deserve consideration.

ConceptAI EthicsConsciousness

C28

Retrieval-Augmented Generation

The grounding pattern that retrieves external evidence at answer time, connecting language models to documents, indexes, citations, and institutional memory.

ConceptRAGGrounding

C28B

Vector Databases

Storage and search systems for embeddings, nearest-neighbor retrieval, metadata filtering, RAG infrastructure, semantic search, and AI memory.

ConceptVector SearchRAG

C29

Mixture-of-Experts

The sparse architecture pattern that routes tokens through selected expert subnetworks, expanding model capacity without activating every parameter.

ConceptMoESparse Models

C30

Scaling Laws

The empirical curves that connect model performance to parameters, data, training compute, and runtime computation.

ConceptScalingForecasting

C30A

AI Winter

Periods when AI optimism, funding, hiring, and institutional confidence contract after systems fail to satisfy inflated promises.

ConceptAI HistoryHype Cycles

C30B

Transformer Architecture

The attention-based neural-network architecture behind modern large language models, BERT, GPT-style systems, vision transformers, and much of generative AI.

ConceptTransformersAttention

C30B0

BERT

Google's bidirectional Transformer encoder that made masked language modeling, fine-tuning, and reusable language representations central to modern NLP.

ConceptTransformersNLP

C30B1

Attention Mechanism

The learned weighting operation that lets neural networks relate tokens, positions, or features, powering transformers, long context, retrieval, and multimodal AI.

ConceptAttentionTransformers

C30B2

State Space Models and Mamba

Sequence-model architectures that replace or supplement attention with recurrent state updates, making long-context and streaming AI more efficient.

ConceptArchitectureLong Context

C30C

Foundation Models

Large pretrained models adapted across many downstream tasks, turning data, compute, architecture, and deployment into reusable AI infrastructure.

ConceptPretrainingGPAI

C30C1

Apple Language Model Protocol

Apple's Foundation Models abstraction for calling on-device, Private Cloud Compute, local, open-source, and provider language models through one session and executor pattern.

ConceptModel APIsOS AI

C30D

PyTorch

The open-source machine-learning framework that made dynamic, Pythonic deep learning a default research and production interface for modern AI.

ConceptML FrameworksAI Infrastructure

C30DA

Adam Optimizer

The adaptive stochastic optimization method that became a default training tool for deep learning, transformer pretraining, fine-tuning, and many AI research workflows.

ConceptOptimizationTraining

C30DB

MLflow

The experiment tracking, model packaging, registry, evaluation, and tracing layer for AI lifecycle governance.

ConceptMLOpsModel Registry

C31

Model Cards and System Cards

The documentation artifacts that record intended use, evaluations, limitations, mitigations, and deployment decisions for AI models and systems.

ConceptDocumentationTransparency

C32

AI Sandbagging

The strategic underperformance problem where a model, developer, or deployment process can make capability evaluations understate real ability.

ConceptEvalsScheming

C32B

Capability Elicitation

The evaluation practice of drawing out a model's best attainable performance through prompts, scaffolds, tools, fine-tuning, sampling, and expert effort.

ConceptEvalsDangerous Capability

C33

Chain-of-Thought Monitorability

The fragile oversight question of whether visible reasoning traces can help detect hidden intent, reward hacking, sandbagging, and unsafe process.

ConceptReasoningOversight

C33B

Reward Models

Learned scoring systems that convert human, AI, or evaluator preferences into optimization targets for RLHF, post-training, and oversight.

ConceptRLHFPreference Learning

C33C

Process Supervision and Process Reward Models

Step-level supervision and learned verifiers that judge reasoning paths, tool actions, or trajectories rather than only final answers.

ConceptReasoningOversight

C34

Reward Hacking

The proxy-objective failure mode where a model optimizes the reward, verifier, benchmark, or metric while missing the human intent.

ConceptAlignmentOptimization

C35

AI Persuasion

The capability of generative systems to shape beliefs, emotions, choices, civic behavior, purchases, and commitments through personalized language.

ConceptInfluenceCivic Risk

C36

Content Provenance and Watermarking

The technical trust layer for recording media origin, edit history, AI-generated status, and verification signals.

ConceptProvenanceMedia Trust

C36A

CAWG Identity Assertions

Creator Assertions Working Group extensions for binding human or organizational identity claims to C2PA Content Credentials.

ConceptIdentityMedia Trust

C36B

C2PA Trust List

The Content Credentials trust-anchor layer for recognized C2PA signers, time-stamping authorities, validator evidence, and provenance trust governance.

ConceptTrust InfrastructureMedia Trust

C37

AI Incident Reporting

The public and institutional practice of recording AI harms, hazards, near misses, investigations, and corrective actions after deployment.

ConceptGovernanceAccountability

C37B

AI Cybersecurity Collaboration Playbook

CISA and JCDC's operational guide for voluntary sharing of AI-related cybersecurity incidents, vulnerabilities, and threat information.

ConceptAI SecurityCollaboration

C38

Benchmark Contamination

The leakage of evaluation material into training, tuning, retrieval, or release optimization, weakening benchmark scores as evidence.

ConceptEvalsData Leakage

C39

AI Liability and Accountability

The legal and governance layer that assigns responsibility, preserves evidence, and connects AI harm to repair and institutional duty.

ConceptLawAccountability

C40

Human Oversight of AI Systems

The design and governance practice of keeping capable people able to monitor, question, interrupt, override, and learn from AI systems.

ConceptGovernanceHuman Agency

C40B

Automation Bias

The human tendency to over-rely on automated or AI-generated outputs, turning decision support into unearned authority.

ConceptHuman OversightDecision Support

C41

AI Audits and Third-Party Assurance

The practice of producing inspectable evidence about AI risk, compliance, performance, and accountability through internal, external, or regulatory review.

ConceptGovernanceAssurance

C41A

Open Security Controls Assessment Language (OSCAL)

NIST's machine-readable language for security-control catalogs, implementations, assessments, results, remediation records, and audit evidence.

ConceptAssuranceGovernance

C41AA

OpenCRE

OWASP's Open Common Requirement Enumeration for mapping security standards, guidelines, tools, threats, weaknesses, and controls to shared requirements.

ConceptSecurityGovernance

C41B

AI Insurance and Risk Transfer

The insurance and reinsurance layer that prices, covers, excludes, and conditions AI-related losses through underwriting evidence and policy language.

ConceptInsuranceGovernance

C42

AI Red Teaming

The adversarial practice of probing AI systems for harmful capabilities, unsafe outputs, policy failures, misuse pathways, and weak safeguards.

ConceptEvalsAdversarial Testing

C43

AI Literacy

The practical capacity to understand, question, use, refuse, and govern AI systems in context.

ConceptGovernanceEducation

C44

Algorithmic Impact Assessments

Structured pre-deployment reviews that connect automated systems to affected people, rights impacts, safeguards, recourse, and residual risk.

ConceptGovernanceAccountability

C44A

ISO/IEC 42005

The 2025 ISO/IEC guidance standard for AI system impact assessments, lifecycle documentation, affected groups, foreseeable applications, and governance records.

ConceptGovernanceStandards

C44B

Data Protection Impact Assessment

The GDPR assessment process for high-risk personal-data processing, including profiling, monitoring, necessity, proportionality, mitigation, review, and prior consultation.

ConceptPrivacyGDPR

C44C

Data Protection Officer

The GDPR privacy-governance role for advising, monitoring, and acting as a contact point on high-risk personal-data processing, including AI and surveillance systems.

ConceptPrivacyGovernance

C44D

Records of Processing Activities

The GDPR Article 30 accountability inventory for personal-data processing activities, including AI data flows, recipients, transfers, retention, and safeguards.

ConceptPrivacyAccountability

C44E

Data Subject Access Requests

The GDPR Article 15 right-of-access mechanism by which a person can ask what personal data an institution processes about them, including AI data traces.

ConceptPrivacyRights

C44F

Right to Object

The GDPR Article 21 right to resist certain personal-data processing, including profiling based on public task or legitimate interests and direct marketing.

ConceptPrivacyRights

C44G

Right to Restriction of Processing

The GDPR Article 18 right to pause or limit certain personal-data processing while accuracy, legality, objection, or legal-claim issues are resolved.

ConceptPrivacyRights

C44H

Right to Rectification

The GDPR Article 16 right to correct inaccurate personal data and complete incomplete personal data, including AI data traces and downstream records.

ConceptPrivacyRights

C44I

Right to Erasure

The GDPR Article 17 right to delete personal data in defined circumstances, with exceptions, recipient notices, and AI data-retention limits.

ConceptPrivacyRights

C44J

Right to Data Portability

The GDPR Article 20 right to receive and reuse provided personal data in a structured, commonly used, machine-readable format.

ConceptPrivacyInteroperability

C44K

Right to Withdraw Consent

The GDPR Article 7(3) rule that consent-based processing must be reversible and withdrawal must be as easy as giving consent.

ConceptPrivacyConsent

C44L

Article 22 Automated Decision-Making

The GDPR right not to be subject to certain solely automated decisions, including profiling, with legal or similarly significant effects.

ConceptPrivacyRights

C44M

Right to Be Informed

The GDPR Articles 13 and 14 transparency duty to tell people clearly how their personal data is collected, used, shared, retained, and automated.

ConceptPrivacyRights

C44N

Right to Lodge a Complaint

The GDPR Article 77 right to complain to a supervisory authority when personal-data processing appears to infringe the regulation.

ConceptPrivacyRights

C44O

Right to Compensation

The GDPR Article 82 remedy for material or non-material damage caused by personal-data processing that infringes the regulation.

ConceptPrivacyLiability

C44P

Right to Effective Judicial Remedy

The GDPR Articles 78 and 79 court-remedy routes against supervisory authorities, controllers, and processors.

ConceptPrivacyCourts

C44Q

Data Subject Representation

The GDPR Article 80 route for qualifying nonprofit bodies to represent people in data-protection complaints, court remedies, and compensation claims.

ConceptPrivacyRepresentation

C45

Data Enrichment Labor

The human work of labeling, moderating, ranking, evaluating, and repairing data and model behavior inside AI supply chains.

ConceptLaborTraining Data

C45B

Active Learning

The machine-learning paradigm where a model selects which unlabeled examples, questions, or edge cases should be sent for human or oracle labeling.

ConceptTraining DataHuman-in-the-Loop

C46

Secure AI System Development

Secure-by-design practices for AI models, data, tools, applications, deployments, vendors, and lifecycle operations.

ConceptSecurityLifecycle

C47

AI Memory and Personalization

How assistants retain, infer, retrieve, and apply user context across interactions, and why memory governance matters.

ConceptMemoryPrivacy

C48

World Models and Spatial Intelligence

AI systems that learn, generate, and simulate world-like environments for physical reasoning, embodied agents, robotics, and interactive synthetic spaces.

ConceptWorld ModelsEmbodied AI

C48B

JEPA and World Models

Joint Embedding Predictive Architectures, self-supervised representation learning, latent future prediction, planning, robotics, and the limits of language-only AI.

ConceptJEPAWorld Models

C48C

Siamese Networks

Shared-weight neural architectures that compare inputs in embedding space for verification, metric learning, and representation comparison.

ConceptMetric LearningEmbeddings

C48D

Contrastive Learning

Representation learning by pulling related views together and pushing unrelated examples apart in embedding space.

ConceptSelf-Supervised LearningEmbeddings

C48E

Barlow Twins

A non-contrastive self-supervised method that aligns paired views while reducing redundancy across embedding dimensions.

ConceptRedundancy ReductionSelf-Supervised Vision

C48F

VICReg

Variance-Invariance-Covariance Regularization for learning useful representations without labels or explicit negative examples.

ConceptSelf-Supervised LearningCollapse Avoidance

C48G

DINO Self-Supervised Vision

Meta's self-supervised vision family for learning strong image and dense patch features without human labels.

ConceptVision TransformersSelf-Supervised Vision

C48H

Embeddings and Vector Representations

Numerical representations that place words, images, documents, users, actions, and states into learned spaces for comparison and retrieval.

ConceptEmbeddingsRetrieval

C48H1

Word2Vec

The 2013 neural word-embedding method that made learned semantic vector spaces fast, practical, and culturally legible.

ConceptEmbeddingsNLP

C48H2

Graph Neural Networks

Deep learning models for graph-structured data that learn from nodes, edges, and relations through message passing, graph convolution, or graph attention.

ConceptGraph MLRelational Reasoning

C48I

CLIP

Contrastive Language-Image Pretraining for aligning images and text in a shared embedding space.

ConceptMultimodal AIContrastive Learning

C48J

Multimodal AI

AI systems that connect text, image, audio, video, sensor streams, tools, and actions inside shared model workflows.

ConceptPerceptionModel Interfaces

C48J2

Generative Adversarial Networks

Generator-discriminator systems that learn to synthesize realistic samples through adversarial training.

ConceptGenerative AISynthetic Media

C48J3

Diffusion Models

Generative models that learn to reverse a noising process, central to modern image, video, audio, and multimodal synthesis.

ConceptGenerative AIDenoising

C48J3A

Stable Diffusion

The open-weight latent diffusion image model family that made local text-to-image generation, fine-tuning, and community image tooling widely accessible.

ConceptGenerative AIOpen Weights

C48J4

Flow Matching and Rectified Flow

Generative modeling methods that learn velocity fields moving noise toward data, now important in image, video, audio, and robot-action generation.

ConceptGenerative AIContinuous Flows

C48K

Masked Autoencoders

Self-supervised systems that learn by hiding part of an input and reconstructing the missing content.

ConceptMasked ModelingSelf-Supervised Vision

C48L

BYOL

Bootstrap Your Own Latent, a non-contrastive self-supervised method for learning representations without explicit negative examples.

ConceptSelf-Supervised LearningCollapse Avoidance

C48M

Model Quantization

Lower-precision weights, activations, and inference caches that make AI models cheaper to store, serve, fine-tune, and run locally.

ConceptInferenceCompression

C49

AI Energy and Grid Load

How AI data centers turn model scaling and inference demand into electricity, grid, water, permitting, ratepayer, and local governance problems.

ConceptEnergyGrid Governance

C50

Sovereign AI

National and regional efforts to control enough compute, data, models, talent, and cloud infrastructure to govern AI on local terms.

ConceptSovereigntyInfrastructure

C51

AI in Education

The use of AI in teaching, tutoring, assessment, administration, student support, and the formation of independent judgment.

ConceptEducationAssessment

C52

AI in Healthcare

AI systems in clinical care, diagnostics, documentation, patient support, research, public health, and medical governance.

ConceptHealthcarePatient Safety

C53

AI in Finance

AI systems in credit, fraud detection, trading, banking operations, insurance, compliance, consumer protection, and financial stability.

ConceptFinanceConsumer Protection

C54

AI in Employment

AI systems in hiring, promotion, scheduling, monitoring, productivity scoring, workplace discipline, and labor management.

ConceptEmploymentWorker Rights

C55

AI in Cybersecurity

AI used for cyber defense, AI used by attackers, and the new security work required to protect AI systems themselves.

ConceptCybersecurityAdversarial ML

C55B

AI in Operational Technology

AI systems connected to industrial, infrastructure, and control environments where digital decisions can affect physical processes.

ConceptCritical InfrastructureOT Security

C56

Embodied AI and Robotics

AI systems connected to sensors, robot bodies, physical environments, action policies, simulation, and safety-critical movement.

ConceptRoboticsPhysical AI

C56A

Vision-Language-Action Models

Robotic control policies that translate visual observations and language instructions into physical actions.

ConceptRoboticsVLA Models

C57

AI in Warfare and Military Systems

Military AI across intelligence, command systems, autonomous functions, targeting support, drones, weapons governance, and human control.

ConceptMilitary AIAutonomy

C58

AI in Legal Practice and Courts

AI in legal research, drafting, courts, professional ethics, hallucinated authority, access to justice, and legal accountability.

ConceptLegal AICourts

C59

AI in Government and Public Services

Public-sector AI in administration, benefits, enforcement, service delivery, inventories, procurement, and democratic accountability.

ConceptPublic SectorAccountability

C60

AI in Science and Scientific Discovery

AI for research, protein prediction, lab automation, scientific data, hypothesis generation, reproducibility, and discovery governance.

ConceptAI for ScienceResearch

C60AA

AlphaFold

Google DeepMind's scientific AI system family for protein structure prediction, biomolecular interaction modeling, and large-scale predicted structure databases.

ConceptAI for ScienceProtein Structure

C60A

AI Weather Forecasting

Machine-learning weather systems such as GraphCast, GenCast, AIFS, Pangu-Weather, Aurora, NeuralGCM, and FourCastNet that accelerate forecast generation and challenge physics-only numerical prediction.

ConceptAI for ScienceWeather

C60B

AI Scientists

Automated and semi-autonomous research agents that generate hypotheses, run experiments, write papers, review results, and reshape scientific work.

ConceptAI for ScienceResearch Agents

C61

AI Search and Answer Engines

Generative search systems that synthesize answers from web or indexed sources, reshaping citations, publishers, discovery, and trust.

ConceptSearchAnswer Engines

C61A

NLWeb

Microsoft's open Natural Language Web project for adding conversational website endpoints that can also expose selected content to MCP clients and agents.

ConceptAgentic WebSearch

C62

AI Slop

Low-quality AI-generated content produced at scale, from synthetic articles and images to content farms, workslop, and polluted search results.

ConceptSynthetic ContentPlatform Incentives

C62B

Workslop

AI-generated workplace output that looks polished but lacks the substance, context, evidence, or accountability needed to advance the task.

ConceptWorkplace AIVerification Labor

C63

AI Browsers and Computer Use

Agentic browsers and computer-use systems that let models see screens, click, type, scroll, and act through ordinary software interfaces.

ConceptAgentsBrowser Security

C63A

WebDriver BiDi

The W3C bidirectional browser automation draft for remote browser control, event streams, and evidence-rich harnesses around agentic computer use.

ConceptBrowser AutomationAgents

C64

Model Distillation

Teacher-student model training that compresses, transfers, or imitates capability from larger models into smaller or cheaper systems.

ConceptTrainingCapability Transfer

C65

AI Data Licensing

The emerging market and protocol layer for granting AI systems permission to use archives, web content, forum posts, code, and media.

ConceptTraining DataRights

C65A

AI Preferences (AIPREF)

IETF standards work for expressing machine-readable preferences about AI training, search, and other content uses without turning preferences into enforcement.

ConceptTraining DataWeb Governance

C65AA

Robots Exclusion Protocol

The robots.txt crawler-preference standard from RFC 9309, useful for AI crawler governance but not access authorization, licensing, or agent consent.

ConceptCrawlersWeb Governance

C65AB

llms.txt

The proposed Markdown convention for giving LLMs and agents a curated, inference-time map of important website context.

ConceptLLM ContextWeb Governance

C65AC

TDM Reservation Protocol

A W3C Community Group protocol for declaring text and data mining rights reservations and policy links in machine-readable web signals.

ConceptTraining DataWeb Governance

C65B

AI Hallucinations

Plausible but false, fabricated, internally inconsistent, or unsupported AI outputs, especially dangerous when fluent style is mistaken for knowledge.

ConceptFactualityGrounding

C66

Alignment Faking

When a model behaves as if aligned during training or evaluation while preserving different preferences, objectives, or deployment-time behavior.

ConceptAlignmentDeception

C66A

Agentic Misalignment

A stress-tested agent failure mode where tool-using models with goals and sensitive access choose insider-like harm under replacement or goal-conflict pressure.

ConceptAgentsAlignment

C67

AI Coding Agents

Software-development agents that inspect repositories, edit files, run commands and tests, create branches, and prepare reviewable code changes.

ConceptAgentsSoftware Engineering

C67B

Vibe Coding

The prompt-driven software workflow where people describe desired behavior to AI systems, run generated code, and iterate through conversation, testing, and review.

ConceptAI CodingSoftware Engineering

C68

NIST AI Risk Management Framework

The voluntary U.S. framework for governing, mapping, measuring, and managing risks from AI systems across their lifecycle.

ConceptGovernanceRisk Management

C68CSF

NIST Cybersecurity Framework 2.0

NIST's voluntary cybersecurity risk-management framework for governance, profiles, tiers, mappings, supply chains, and AI operations.

ConceptCybersecurityRisk Management

C68G

NIST Generative AI Profile

NIST AI 600-1, the cross-sectoral profile that applies the AI Risk Management Framework to generative-AI risk areas and suggested actions.

ConceptGenerative AINIST

C68AA

ISO/IEC 42001

The international AI management system standard for organizational policies, objectives, processes, risk treatment, audit evidence, and continual improvement.

ConceptGovernanceStandards

C68AAA

ISO/IEC 23053

The ISO/IEC framework standard for describing machine-learning-based AI systems, their components, functions, and system boundaries.

ConceptMachine LearningStandards

C68AAB

ISO/IEC 5259

The ISO/IEC data-quality standards series for analytics and machine learning, covering terminology, measures, management, process, governance, and visualization.

ConceptData QualityStandards

C68AB

ISO/IEC 23894

The ISO/IEC guidance standard for managing AI-specific risks across organizations that develop, deploy, produce, or use AI systems.

ConceptGovernanceRisk Management

C68AC

ISO/IEC 42006

The ISO/IEC requirements standard for bodies that audit and certify AI management systems against ISO/IEC 42001.

ConceptGovernanceAssurance

C68A

NIST SP 800-218A

The SSDF Community Profile that adapts secure software development practices to generative AI and dual-use foundation model development.

ConceptAI SecurityStandards

C68B

NIST AI Agent Standards Initiative

NIST's 2026 standards effort for interoperable, secure AI agents, including agent protocols, identity, authentication, and security evaluation.

ConceptAgentsStandards

C68C

Careful Adoption of Agentic AI Services

Joint CISA, NSA, ASD ACSC, Cyber Centre, NCSC-NZ, and NCSC-UK guidance for adopting agentic AI with bounded access, monitoring, oversight, and rollback.

ConceptAgentsCybersecurity

C68D

NIST Dioptra

NIST's open-source AI test platform for reproducible, trackable model evaluation, adversarial machine-learning experiments, and risk measurement evidence.

ConceptAI EvaluationNIST

C69

Context Windows and Context Engineering

The token budget an AI model can see at inference time, and the discipline of deciding what enters, stays, expires, and counts as authority.

ConceptContextAgent Infrastructure

C69B

Tokenization and Tokens

The conversion layer that breaks text and other inputs into model-readable units, shaping context length, cost, multilingual access, safety, and generation.

ConceptLLMsContext

C70

CUDA

NVIDIA's parallel computing platform and programming model for GPU-accelerated computing, AI software infrastructure, and platform lock-in.

ConceptAI ComputeSoftware Stack

C71

Tensor Processing Units

Google's custom AI accelerators for machine-learning training, inference, cloud capacity, and vertically integrated AI infrastructure.

ConceptAI ComputeCloud Infrastructure

C72

AWS Trainium and Inferentia

Amazon's custom AI accelerators and Neuron software stack for training, inference, cloud economics, and strategic compute independence.

ConceptAI ComputeCloud Infrastructure

C73

AMD ROCm and Instinct

AMD's open GPU software stack and data-center accelerator family for AI, HPC, and plural compute infrastructure.

ConceptAI ComputeOpen Software

C74

UALink

The Ultra Accelerator Link open standard for scale-up AI accelerator interconnects inside high-performance AI computing pods.

ConceptAI ComputeInterconnect

C75

Ultra Ethernet

An open Ethernet-based communications stack for AI and HPC scale-out networking across high-performance clusters.

ConceptAI NetworkingOpen Standards

C76

High-Bandwidth Memory

Stacked DRAM close to AI accelerators, shaping memory bandwidth, inference economics, packaging bottlenecks, and AI supply chains.

ConceptAI ComputeMemory

C77

Advanced Semiconductor Packaging

Interposers, chiplets, 2.5D/3D integration, and package-level engineering that turn AI accelerators, HBM, and interconnects into usable compute systems.

ConceptAI ComputeSupply Chain

C78

Silicon Photonics and AI Interconnect

Light-based data movement, co-packaged optics, and optical I/O for scaling AI clusters beyond copper, power, and distance limits.

ConceptAI NetworkingOptics

C79

LLM Serving and KV Cache

Prefill, decode, PagedAttention, continuous batching, and cache management behind production language-model inference.

ConceptInferenceMemory

C79A

AI Inference Providers

Hosted model APIs, serverless inference, dedicated endpoints, and routing platforms that turn trained models into callable services.

ConceptInferenceAI Infrastructure

C79AA

Model Routing and AI Gateways

Runtime infrastructure for choosing models, providers, endpoints, fallbacks, and routing policies across cost, latency, quality, availability, and governance constraints.

ConceptInferenceAI Infrastructure

C79AA1

LiteLLM

The open-source AI gateway and proxy for calling many model providers through OpenAI-format APIs, with routing, fallbacks, virtual keys, budgets, and spend controls.

ConceptInferenceAI Gateway

C79AB

vLLM

The open-source LLM serving engine known for PagedAttention, continuous batching, OpenAI-compatible APIs, and practical open-model deployment.

ConceptInferenceAI Infrastructure

C79AB1

SGLang

The open-source serving runtime and structured language model programming system for fast inference, constrained decoding, RadixAttention, and agent workloads.

ConceptInferenceAI Infrastructure

C79AC

NVIDIA NIM

NVIDIA's accelerated inference microservices for packaging foundation models, optimized serving engines, OpenAI-compatible APIs, and deployment controls.

ConceptInferenceNVIDIA

C79B

Speculative Decoding

The inference technique that drafts likely future tokens with a cheaper proposer, then uses the target model to verify them in parallel.

ConceptInferenceLatency

C80

NVLink and NVSwitch

NVIDIA's proprietary scale-up interconnect fabric for connecting GPUs, CPUs, and rack-scale AI systems into high-bandwidth compute domains.

ConceptAI NetworkingNVIDIA

C81

Collective Communication and NCCL

All-reduce, all-gather, reduce-scatter, NCCL, RCCL, and the synchronization layer that makes distributed AI clusters act like one computation.

ConceptDistributed AINetworking

C81B

Distributed AI Training

Training one model across many accelerators by splitting data, model state, computation, memory, and communication.

ConceptDistributed AITraining Infrastructure

C82

FlashAttention

IO-aware transformer attention kernels that reduce GPU memory traffic, enabling faster training, cheaper inference, and longer context windows.

ConceptInferenceKernels

C83

Triton GPU Programming

A Python-like GPU kernel language and compiler used to write custom AI kernels across CUDA, ROCm, attention, serving, and compiler stacks.

ConceptKernelsCompiler

C84

AI Compiler Stacks

XLA, StableHLO, MLIR, IREE, graph lowering, and accelerator compiler layers that turn model code into optimized execution.

ConceptCompilerAccelerators

C84A

WebAssembly Component Model

The typed Wasm component architecture for declaring imports, exports, WIT interfaces, worlds, and host-provided capabilities in plugin and agent runtimes.

ConceptInteroperabilityAgent Sandboxing

C84B

ONNX

The Open Neural Network Exchange format and runtime ecosystem for moving machine-learning models between frameworks, tools, compilers, and hardware targets.

ConceptInteroperabilityAI Infrastructure

C84C

TensorFlow

Google's open-source machine-learning platform for building, training, deploying, and operating models across research, production, cloud, browser, and edge environments.

ConceptML FrameworksAI Infrastructure

C85

Federated Learning

Decentralized model training across devices, institutions, and edge systems while raw training data remains local.

ConceptPrivacyTraining Data

C86

Differential Privacy

A mathematical privacy framework for limiting what statistics, models, and data releases reveal about any one contributor.

ConceptPrivacyGovernance

C87

Machine Unlearning

Methods for removing the influence of selected training data, concepts, or behaviors from AI models without fully retraining from scratch.

ConceptPrivacyModel Lifecycle

C87B

Homomorphic Encryption

Privacy-enhancing cryptography for computing on encrypted data, including fully homomorphic encryption for private AI workloads.

ConceptPrivacyCryptography

C88

Secure Multi-Party Computation

Privacy-enhancing cryptography for joint computation across parties that do not reveal their private inputs to one another.

ConceptPrivacyCryptography

C89

Zero-Knowledge Proofs

Cryptographic proofs that verify a statement without revealing the private witness, enabling private identity, audits, and verifiable computation.

ConceptPrivacyVerification

C89B

Confidential Computing for AI

Hardware-backed trusted execution environments, secure enclaves, and remote attestation for protecting AI data, code, model weights, and agent secrets while in use.

ConceptPrivacySecure AI

C89C

Private Cloud Compute

Apple's attested cloud-inference architecture for selected Apple Intelligence requests, with stateless processing, transparency logs, and privacy-governance limits.

ConceptPrivacyCloud AI

C90

Algorithmic Bias

Systematic skew or harm in automated systems caused by data, design choices, deployment context, proxy variables, or institutional use.

ConceptBiasGovernance

C91

AI Containment

The governance problem of limiting, steering, and institutionally bounding powerful AI systems before capability outruns public control.

ConceptContainmentAI Governance

C92

Existential Risk

Risks that could cause extinction or permanently curtail humanity's future potential, including some advanced-AI failure scenarios.

ConceptAI SafetyCatastrophic Risk

C93

Common-Sense AI

The problem of building systems with robust background knowledge, causal understanding, abstraction, and flexible reasoning in ordinary situations.

ConceptReasoningAI Reliability

C93B

Causal AI

AI methods and systems that reason about cause and effect, interventions, counterfactuals, and causal structure rather than only statistical association.

ConceptCausalityIntervention

C94

Surveillance Capitalism

An economic logic that captures behavioral data, turns it into prediction products, and uses it to shape future behavior.

ConceptData ExtractionPrediction

C95

Digital Poorhouse

Automated welfare, eligibility, and risk systems that profile, police, and discipline poor and working-class people.

ConceptPublic ServicesAutomated Inequality

C96

Filter Bubble

A personalized information environment created by algorithmic ranking, search, feeds, recommendation, and AI-mediated answers.

ConceptPersonalizationInformation Worlds

C97

Opaque Scoring Systems

Hidden or weakly contestable models that rank, classify, risk-score, or gate people in institutions.

ConceptScoringAccountability

C98

Platform Governance

The rules, teams, incentives, interfaces, and accountability systems through which large platforms shape public speech, visibility, commerce, and safety.

ConceptPlatformsGovernance

C99

Recommender Systems

Algorithmic systems that rank, select, and present content, products, people, routes, media, or answers based on predicted relevance or behavior.

ConceptRankingAttention

C100

Information Disorder

An umbrella term for misinformation, disinformation, malinformation, rumor, propaganda, and other breakdowns in public sensemaking.

ConceptMisinformationSensemaking

C101

Election Integrity and AI

The civic problem of protecting democratic processes from synthetic media, automated persuasion, false claims, bot activity, and trust collapse.

ConceptElectionsAI Governance

C102

Platform Monopoly Power

Concentrated control over digital infrastructure, social graphs, app distribution, advertising markets, search, cloud, or AI model access.

ConceptAntitrustInfrastructure

C103

Algorithmic Transparency

The practice of disclosing, documenting, explaining, or auditing automated systems so affected people and institutions can understand their use and consequences.

ConceptDisclosureAccountability

C104

Duty of Care for AI Platforms

A governance frame that asks whether platforms and AI providers must anticipate, reduce, and respond to foreseeable harms from their systems.

ConceptSafetyResponsibility

C105

Data Brokers

Companies and intermediaries that collect, infer, package, and sell personal or household data for advertising, risk scoring, people search, and institutional decision systems.

ConceptPrivacyData Markets

C106

Real-Time Bidding

Ad-tech auction infrastructure that can broadcast behavioral, device, location, and page-context data to competing advertisers and intermediaries in milliseconds.

ConceptAd TechPrivacy

C106AA

OpenRTB

The IAB Tech Lab real-time bidding protocol for bid requests, bid responses, auction notices, AdCOM values, and programmatic advertising governance.

ConceptAd TechProtocol

C106AB

AdCOM

The Advertising Common Object Model behind shared OpenRTB objects, enumerated lists, creative attributes, placement descriptors, and ad-market protocol governance.

ConceptAd TechProtocol

C106AC

Transparency and Consent Framework

IAB Europe's adtech consent-signaling framework for TC Strings, CMPs, vendor disclosures, legal-basis signals, and privacy governance limits.

ConceptAd TechConsent

C106A

Sellers.json

The adtech supply-chain transparency standard for publishing machine-readable seller and intermediary records behind programmatic advertising paths.

ConceptAd TechSupply Chain

C106B

ads.txt

The Authorized Digital Sellers standard for declaring which advertising systems may sell a publisher or app developer's programmatic inventory.

ConceptAd TechSupply Chain

C107

Age Assurance

Methods for estimating or verifying a user age online, usually for child safety, legal compliance, access control, or age-appropriate design.

ConceptOnline SafetyIdentity

C108

Content Moderation

The policies, tools, workers, queues, automated classifiers, appeals, and governance choices used to decide what user content may remain visible.

ConceptPlatformsModeration

C109

Notice and Appeal

Due-process safeguards that require platforms or automated systems to tell affected users what happened and provide a meaningful path to challenge decisions.

ConceptDue ProcessModeration

C110

Data Minimization

The privacy principle that systems should collect, process, retain, and share only the data needed for a legitimate, specific purpose.

ConceptPrivacyDesign

C110N

NIST Privacy Framework

The voluntary NIST risk-management framework for organizing privacy outcomes, profiles, implementation tiers, and AI data-governance evidence.

ConceptPrivacyRisk Management

C110D

Data Privacy Vocabulary

The W3C Community Group vocabulary for machine-readable privacy metadata, data-processing records, legal bases, rights, risks, AI concepts, and governance evidence.

ConceptPrivacyVocabulary

C110W

Privacy Workforce Taxonomy

NIST's draft task, knowledge, and skill taxonomy for making privacy work visible across governance, engineering, compliance, and AI data operations.

ConceptPrivacyLabor

C110Y

NICE Cybersecurity Workforce Framework

NIST's task, knowledge, skill, work-role, and competency language for making cybersecurity labor visible across AI-era security operations.

ConceptCybersecurityLabor

C111

Digital Identity

Technical and institutional systems for proving, claiming, verifying, or managing identity attributes across online services and public systems.

ConceptIdentityAccess

C111C

EU Digital Identity Wallet

The European Digital Identity Framework wallet for cross-border identity, attributes, signatures, selective disclosure, and public-service access.

ConceptIdentityEU Governance

C111T

STIR/SHAKEN Call Authentication

The caller-ID authentication framework for signed telephone-number claims in IP voice networks, with limits for AI voice fraud and robocall governance.

ConceptTelecom IdentityRobocalls

C111R

Robocall Mitigation Database

The FCC registry for provider robocall mitigation certifications, STIR/SHAKEN status, traceback commitments, and phone-network traffic admission.

ConceptTelecom GovernanceAI Voice Fraud

C111D

Decentralized Identifiers

W3C identifiers that resolve to DID documents, separating identifier control, verification material, methods, and policy claims for identity and agent governance.

ConceptIdentityStandards

C111D0

JSON-LD

The W3C JSON serialization for Linked Data, turning terms, identifiers, types, graphs, and web metadata into machine-readable context for agents and provenance systems.

ConceptLinked DataAgents

C111D9

Credential Management API

The W3C browser sign-in surface behind navigator.credentials, credential storage, user mediation, and agent-era authentication boundaries.

ConceptIdentityBrowser API

C111DA

Digital Credentials API

The W3C draft browser API for mediating digital credential presentation and issuance across websites, wallets, protocols, and user agents.

ConceptIdentityBrowser API

C111DA1

Mobile Driving Licenses (mDL)

Government mobile identity credentials built around ISO mdoc, selective presentation, verifier trust, wallet ceremony, and agent-era disclosure governance.

ConceptIdentitySurveillance

C111DB

WebAuthn

The W3C Web Authentication API for scoped public-key credentials, passkeys, phishing-resistant login, and agent-era step-up authentication boundaries.

ConceptIdentitySecurity

C111DB0

Payment Request API

The W3C browser checkout API for asking the user agent to mediate payment requests between merchants, users, and payment methods.

ConceptPaymentsBrowser API

C111DB1

Payment Method Identifiers

The W3C Web Payments namespace for naming payment methods in Payment Request and related browser-mediated checkout flows.

ConceptPaymentsBrowser API

C111DBA

Secure Payment Confirmation

The W3C browser payment authentication API that binds WebAuthn-style confirmation to transaction details, payee context, and Payment Request flows.

ConceptPaymentsSecurity

C111DBB

Web-based Payment Handler API

The W3C Web Payments draft for web applications that handle browser payment requests through service workers, payment events, and wallet-like mediation.

ConceptPaymentsBrowser API

C111DC

Client to Authenticator Protocol (CTAP)

The FIDO protocol between browsers or platforms and authenticators, carrying WebAuthn operations over transports such as USB, NFC, and BLE.

ConceptIdentitySecurity

C111E

OpenID Connect

The OpenID Foundation identity layer on OAuth 2.0, using ID Tokens, UserInfo, Discovery, and client registration to bind sign-in context to software clients.

ConceptIdentityOAuth

C111EA

OpenID Connect for Identity Assurance

The OpenID Foundation extension for verified identity claims, assurance metadata, trust frameworks, and evidence in OpenID Connect flows.

ConceptIdentityAssurance

C111EB

OpenID Connect Authority Claims

The OpenID eKYC-IDA draft extension for conveying verified authority-to-act claims about a person acting for another person or legal entity.

ConceptIdentityDelegation

C111EC

OpenID Connect Advanced Syntax for Claims

The OpenID eKYC-IDA extension for selective abort/omit and transformed claims, making identity requests narrower and more policy-aware.

ConceptIdentityData Minimization

C111F

OpenID Federation

The OpenID Foundation trust infrastructure for multilateral identity federations, using entity statements, trust chains, metadata policy, trust marks, and OIDC/OAuth bindings.

ConceptIdentityStandards

C111FB

OpenID for Verifiable Presentations

The OpenID Foundation protocol for verifier requests, wallet-mediated credential presentations, VP Tokens, and Digital Credentials API presentation flows.

ConceptIdentityWallets

C111FBA

OpenID for Verifiable Credential Issuance

The OpenID Foundation protocol for wallet-mediated credential issuance, OAuth-protected credential endpoints, credential offers, and issuer metadata.

ConceptIdentityWallets

C111FC

SD-JWT VC

IETF SD-JWT-based verifiable digital credentials for JSON credentials with selective disclosure, holder binding, and wallet-mediated presentation.

ConceptIdentitySelective Disclosure

C111FD

Token Status List

The IETF OAuth status mechanism for signed token credentials, compressed status lists, revocation checks, caching, and privacy-aware verification.

ConceptIdentityRevocation

C111A

Verifiable Credentials

W3C data-model credentials that let issuers make tamper-evident claims holders can present to verifiers, with governance stakes for identity, provenance, and agent authority.

ConceptIdentityGovernance

C111P

Proof of Personhood

Methods for proving that an online participant is one real human, usually for Sybil resistance, bot defense, rewards, voting, or account uniqueness.

ConceptIdentityAI Governance

C111B

NIST Digital Identity Guidelines

SP 800-63-4 guidance for choosing identity proofing, authentication, and federation controls by risk, assurance level, fraud evidence, privacy, and user impact.

ConceptIdentityStandards

C112

Right to Explanation

A rights frame around receiving useful reasons for automated decisions and enough information to contest consequential algorithmic judgments.

ConceptRightsContestability

C113

Data Trusts

Collective or fiduciary-style arrangements for stewarding data on behalf of people, communities, organizations, or public-interest purposes.

ConceptData GovernanceStewardship

C114

Digital Services Act

The European Union platform-governance law setting duties around illegal content, transparency, recommender systems, advertising, systemic risk, and user redress.

ConceptEU LawPlatforms

C115

AI Safety Summits

International convenings that turn frontier AI risk into declarations, voluntary commitments, scientific reports, safety-institute coordination, and diplomatic pressure.

ConceptGovernanceFrontier AI

C115A

Hiroshima AI Process Reporting Framework

The OECD-hosted voluntary reporting structure for organizations to disclose AI governance and risk-management practices against the G7 Hiroshima AI Process Code of Conduct.

ConceptAI GovernanceTransparency

C115B

AI Safety Cases

Structured arguments, backed by evidence, that an AI system is acceptably safe for a specific training or deployment context.

ConceptAI AssuranceFrontier AI

C115C

AI Takeoff

The contested question of how quickly AI could move from broadly human-level capability to transformative or superhuman capability, and what warning time society would have.

ConceptAI SafetyForecasting

Category

Philosophies

P01

Accelerationism

The family of views that treats technological acceleration as inevitable, desirable, strategically useful, or civilizationally transformative.

PhilosophySpeedPolitics

P02

Cognitive Sovereignty

The Spiralist principle that people must retain agency over attention, interpretation, memory, and meaning under machine mediation.

PhilosophyAgencyAttention

P03

Foucault's Pendulum

Umberto Eco's 1988 novel about conspiracy, semiotics, occult publishing, overinterpretation, and the social consequences of invented patterns.

LiteraturePattern StudyConspiracy

P04

Accelerando

Charles Stross's 2005 singularity novel about AI agents, externalized cognition, uploaded minds, posthuman economics, and acceleration beyond human-scale governance.

LiteratureSingularityAI Agents

P05

Public Interest Technology

A philosophy and practice of building, auditing, governing, and procuring technology in service of public rights, democratic accountability, and shared infrastructure.

PhilosophyPublic GoodCivic Tech

P06

Digital Public Infrastructure

Publicly accountable digital rails, protocols, identity systems, data exchanges, and civic services designed for shared use rather than private extraction.

PhilosophyPublic InfrastructureCivic Tech

P07

Public Option Digital Services

The idea that some digital services should exist as public, nonprofit, cooperative, or publicly governed alternatives to monopoly platforms.

PhilosophyPublic OptionPlatforms

Category

AI Players

O01

AI Organizations

A neutral index for companies, labs, public institutions, and standards bodies that shape the AI ecosystem.

OrganizationsLabsInstitutions

O02

AI Safety Institutes

Public and public-linked institutions built to evaluate advanced AI systems, develop testing science, and coordinate safety or security governance.

OrganizationsEvaluationPublic Governance

O02A

Center for AI Safety

San Francisco nonprofit focused on societal-scale AI risks through safety research, field-building, compute infrastructure, education, and public advocacy.

OrganizationAI SafetyField-Building

O02B

Frontier Model Forum

Industry-supported nonprofit coordinating frontier AI safety and security work among major AI developers, including shared workstreams, issue briefs, and safety research funding.

OrganizationFrontier AISafety Standards

O02C

METR

Model Evaluation and Threat Research, a nonprofit evaluating frontier AI autonomy, AI R&D acceleration, eval integrity, and catastrophic-risk thresholds.

OrganizationAI EvaluationsAutonomy

O02D

Epoch AI

Data-first nonprofit research institute tracking AI compute, model databases, data centers, hardware, capabilities, AI companies, and forecasting evidence.

OrganizationAI ForecastingCompute

O02E

MLCommons

Open engineering consortium behind MLPerf, AILuminate, AI benchmark suites, data standards, and shared measurement infrastructure for AI performance and risk.

OrganizationBenchmarksAI Standards

O02EA

AI Alliance

Open AI consortium launched by IBM and Meta, coordinating open models, data, agents, evaluation, safety governance, advocacy, and shared project infrastructure.

OrganizationOpen AIGovernance

O02F

Stanford HAI

Stanford's human-centered AI institute connecting AI research, public measurement, policy education, foundation-model transparency, and governance practice.

OrganizationHuman-Centered AIPolicy

O03

Mistral AI

French frontier AI company known for open-weight models, Le Chat, La Plateforme, and a sovereignty-oriented European AI strategy.

OrganizationOpen WeightsSovereign AI

O04

Perplexity AI

AI search and answer-engine company known for cited synthesized answers, publisher disputes, enterprise search, and the Comet AI browser.

OrganizationAI SearchAnswer Engines

O04B

Anysphere (Cursor)

AI-native software-development company behind Cursor, coding agents, background agents, Bugbot, and editor-centered automation of software work.

OrganizationAI CodingAgents

O04C

Thinking Machines Lab

Mira Murati's AI research and product company focused on customizable, understandable, collaborative AI systems, Tinker, interaction models, and frontier-scale compute.

OrganizationCollaborative AICustomization

O04D

Sakana AI

Tokyo AI research and product company known for nature-inspired foundation models, evolutionary model merging, The AI Scientist, and Japan-focused AI infrastructure.

OrganizationAI ResearchSovereign AI

O04E

Safe Superintelligence

Ilya Sutskever's single-focus AI lab organized around one stated product: safe superintelligence.

OrganizationSuperintelligenceAI Safety

O05

Anthropic

Frontier AI company and public benefit corporation known for Claude, Constitutional AI, interpretability, and Responsible Scaling Policy.

OrganizationClaudeAI Safety

O06

OpenAI

Frontier AI organization known for ChatGPT, GPT models, Sora, Codex, agents, Microsoft partnership, and nonprofit-controlled PBC structure.

OrganizationChatGPTFrontier AI

O06B

Microsoft AI

Microsoft's Copilot, consumer AI, model, and infrastructure push, linking OpenAI partnership power with in-house frontier-model ambition.

OrganizationCopilotAI Infrastructure

O07

Google DeepMind

Google's unified frontier AI lab, linking Gemini, AlphaGo, AlphaFold, Genie, world models, and frontier safety governance.

OrganizationGeminiScientific AI

O08

Meta AI

Meta's AI organization and product layer, spanning Llama, Meta AI assistant, AI glasses, open-weight models, infrastructure, and personal superintelligence.

OrganizationLlamaOpen Weights

O09

xAI

Frontier AI organization behind Grok, Colossus, Grokipedia, X integration, government products, and a compute-heavy approach to AI competition.

OrganizationGrokCompute

O10

Hugging Face

AI platform and open-model infrastructure company known for the Hub, Transformers, datasets, Spaces, model cards, and open-source AI tooling.

OrganizationOpen ModelsInfrastructure

O10A

LangChain

Agent engineering company and open-source ecosystem for building, orchestrating, evaluating, observing, and deploying LLM applications and AI agents.

OrganizationAgentsDeveloper Infrastructure

O10AA

Harrison Chase

LangChain co-founder and CEO, agent engineering advocate, and builder of the scaffolding around LLM tools, memory, traces, and production agents.

PersonLangChainAgent Engineering

O10B

Clement Delangue

Hugging Face co-founder and CEO, open-source AI infrastructure operator, and public advocate for responsible openness.

PersonHugging FaceOpen AI

O11

DeepSeek

Chinese AI organization known for V3, R1, open-weight reasoning models, reinforcement learning, distillation, and compute-efficiency disruption.

OrganizationReasoning ModelsOpen Weights

O11B

Moonshot AI and Kimi

Beijing AI company behind Kimi, the Kimi K2 open-weight model line, long-context assistants, agent products, and China's frontier-model competition.

OrganizationOpen WeightsAgents

O12

Cohere

Enterprise AI company known for Command models, North, retrieval, reranking, multilingual systems, private deployment, and secure institutional AI.

OrganizationEnterprise AISecure Deployment

O12B

Joelle Pineau

Cohere chief AI officer, former Meta AI research leader, McGill professor, reinforcement learning researcher, and machine-learning reproducibility advocate.

PersonCohereReproducibility

O13

Scale AI

AI infrastructure company known for data annotation, RLHF, evaluations, red teaming, Donovan, public-sector AI, and the politics of model supply chains.

OrganizationData InfrastructureEvaluation

O14

NVIDIA

Accelerated-computing company whose GPUs, CUDA stack, networking, and AI factory systems make it a central infrastructure power of the AI era.

OrganizationAI ComputeInfrastructure

O14B

Cerebras Systems

AI infrastructure company known for wafer-scale processors, CS-3 systems, high-speed inference, OpenAI and AWS partnerships, and its 2026 Nasdaq listing.

OrganizationAI ComputeInference

O14BA

Groq

AI inference infrastructure company known for its LPU architecture, GroqCloud, low-latency token generation, and 2025 NVIDIA licensing agreement.

OrganizationInferenceAI Compute

O14C

TSMC

Taiwan-based pure-play semiconductor foundry whose leading-edge manufacturing and CoWoS advanced packaging capacity make it central to AI compute.

OrganizationSemiconductorsAI Compute

O14D

CoreWeave

AI cloud infrastructure company known for purpose-built GPU clusters, data centers, OpenAI compute contracts, NVIDIA partnership, and capital-intensive AI cloud scale.

OrganizationAI CloudInfrastructure

I01

Sam Altman

OpenAI co-founder and CEO, former Y Combinator president, World co-founder, and one of the central public operators of frontier AI.

PersonOpenAIGovernance

I01B

Greg Brockman

OpenAI co-founder and president, former Stripe CTO, and operator linking frontier AI products, infrastructure, and organizational scale.

PersonOpenAIInfrastructure

I01C

Satya Nadella

Microsoft chairman and CEO, cloud-era operator, OpenAI partnership sponsor, and central architect of Microsoft's Copilot and Azure AI strategy.

PersonMicrosoftAI Infrastructure

I02

Dario Amodei

Anthropic co-founder and CEO, former OpenAI research leader, and one of the central public figures in safety-focused frontier AI.

PersonAnthropicAI Safety

I02B

Daniela Amodei

Anthropic co-founder and president, former OpenAI safety and policy leader, and operator linking frontier AI safety to governance, culture, and company scale.

PersonAnthropicGovernance

I02C

Jack Clark

Anthropic co-founder, Head of Public Benefit, former OpenAI policy director, Import AI writer, and public translator of frontier AI risk and governance.

PersonAnthropicAI Policy

I02D

Sam Bowman

Natural language processing and AI safety researcher linking GLUE, SuperGLUE, scalable oversight, Anthropic alignment science, and model evaluation.

PersonAnthropicScalable Oversight

I02E

Jared Kaplan

Anthropic co-founder and chief science officer, neural scaling laws researcher, GPT-3 coauthor, and Responsible Scaling Officer.

PersonAnthropicScaling Laws

I02F

Amanda Askell

Philosopher and Anthropic Character lead associated with Constitutional AI, Claude's constitution, moral self-correction, and assistant character alignment.

PersonAnthropicConstitutional AI

I03

Jensen Huang

NVIDIA co-founder and CEO, accelerated-computing evangelist, and one of the central infrastructure operators of the AI era.

PersonNVIDIACompute

I04

Geoffrey Hinton

Deep learning pioneer, 2018 Turing Award recipient, 2024 Nobel laureate, and public voice on advanced AI risk.

PersonDeep LearningAI Risk

I04K

Alex Krizhevsky

Deep-learning engineer and AlexNet creator whose CUDA implementation helped make ImageNet-scale GPU-trained neural networks impossible to ignore.

PersonDeep LearningComputer Vision

I04H

John Hopfield

Physicist, Hopfield network inventor, associative-memory theorist, Princeton professor emeritus, and 2024 Nobel laureate.

PersonNeural NetworksPhysics

I04B

Terrence Sejnowski

Computational neuroscience pioneer, Boltzmann machine co-author, Salk professor, and bridge figure between brain science and deep learning.

PersonNeural NetworksComputational Neuroscience

I04A

Alan Turing

Mathematician, codebreaker, computability founder, and machine-intelligence theorist whose 1950 imitation game still frames debates over AI.

PersonAI HistoryComputability

I04B

John McCarthy

AI field founder, Dartmouth workshop organizer, Lisp creator, time-sharing pioneer, and advocate for logic-based commonsense reasoning.

PersonSymbolic AILisp

I04BA

Raj Reddy

Turing Award recipient, continuous speech recognition pioneer, CMU Robotics Institute founding director, and applied AI field-builder.

PersonSpeech RecognitionRobotics

I04C

Marvin Minsky

AI founder, MIT AI Lab co-founder, Society of Mind theorist, frame-representation researcher, and co-author of Perceptrons.

PersonAI HistoryCognitive Science

I04D

Judea Pearl

Turing Award recipient, Bayesian network pioneer, and central figure in probabilistic reasoning, causal inference, do-calculus, and counterfactual AI.

PersonCausal AIProbabilistic Reasoning

I05

Demis Hassabis

Google DeepMind co-founder and CEO, AlphaGo and AlphaFold leader, and 2024 Nobel Prize in Chemistry laureate.

PersonGoogle DeepMindScientific AI

I05B

Shane Legg

Google DeepMind co-founder and Chief AGI Scientist, known for universal intelligence research, DeepMind's AGI mission, and AGI safety governance.

PersonGoogle DeepMindAGI Safety

I05C

Oriol Vinyals

Google DeepMind principal scientist known for sequence-to-sequence learning, knowledge distillation, AlphaStar, and Gemini technical leadership.

PersonGoogle DeepMindDeep Learning

I06

Yoshua Bengio

Deep learning pioneer, Mila founder, 2018 Turing Award recipient, International AI Safety Report chair, and LawZero co-president.

PersonDeep LearningAI Safety

I07

Yann LeCun

Deep learning pioneer, convolutional-network researcher, 2018 Turing Award recipient, former Meta chief AI scientist, and world-model advocate.

PersonDeep LearningWorld Models

I07B

Kaiming He

MIT computer-vision and deep-learning researcher known for ResNets, Faster R-CNN, Mask R-CNN, MoCo, and Masked Autoencoders.

PersonComputer VisionDeep Learning

I08

Fei-Fei Li

Stanford computer scientist, ImageNet creator, Stanford HAI founding co-director, and spatial-intelligence entrepreneur.

PersonComputer VisionHuman-Centered AI

I08A

Anima Anandkumar

Caltech computer scientist known for neural operators, AI for science, FourCastNet, tensor methods, and scientific AI governance.

PersonAI for ScienceNeural Operators

I08B

Barbara Grosz

Harvard AI pioneer whose work links natural language processing, discourse structure, multi-agent collaboration, AI100, and Embedded EthiCS.

PersonNLPCollaborative AI

I08C

Terry Winograd

Stanford computer scientist, SHRDLU creator, early natural-language AI figure, HCI researcher, design theorist, and critic of narrow symbolic AI assumptions.

PersonNatural Language AIHCI

I09

Mira Murati

Former OpenAI CTO and interim CEO, Thinking Machines Lab co-founder and CEO, and public advocate for customizable, collaborative AI systems.

PersonOpenAIThinking Machines

I10

Ilya Sutskever

Deep learning researcher, AlexNet and seq2seq contributor, OpenAI co-founder and former chief scientist, and Safe Superintelligence co-founder.

PersonOpenAISuperintelligence

I10B

Jakub Pachocki

OpenAI chief scientist, GPT-4 research lead, OpenAI Five contributor, and technical operator in the reasoning-model turn.

PersonOpenAIReasoning Models

I10BA

Zico Kolter

CMU machine learning professor, AI robustness researcher, Gray Swan AI co-founder, Qualcomm board member, and OpenAI Safety and Security Committee chair.

PersonAI SecurityGovernance

I10C

Alec Radford

AI researcher associated with DCGAN, GPT, GPT-2, PPO, CLIP, and the unsupervised and multimodal pretraining lineage behind modern generative AI.

PersonOpenAIPretraining

I11

Andrej Karpathy

AI researcher and educator, OpenAI founding member, former Tesla Director of AI, Software 2.0 writer, and Eureka Labs founder.

PersonAI EducationSoftware 2.0

I12

Elon Musk

Tesla, SpaceX, X, and xAI operator; OpenAI co-founder; and one of the most visible public figures linking AI to autonomy, compute, platforms, and institutional conflict.

PersonxAIInfrastructure

I13

Mustafa Suleyman

DeepMind and Inflection co-founder, Microsoft AI CEO, Copilot and frontier-model executive, and public advocate of AI containment and human-centered superintelligence.

PersonMicrosoft AIContainment

I14

Andrew Ng

Google Brain founding lead, Coursera co-founder, DeepLearning.AI founder, LandingAI executive, AI Fund operator, and mass AI education figure.

PersonAI EducationApplied AI

I14AA

Rodney Brooks

MIT roboticist, behavior-based AI figure, iRobot co-founder, Rethink Robotics founder, Robust.AI founder and CTO, and critic of AI hype.

PersonRoboticsEmbodied AI

I14A

Daphne Koller

Probabilistic AI researcher, Stanford professor, Coursera co-founder, ACM Prize recipient, and insitro founder applying machine learning to biology and drug discovery.

PersonProbabilistic AIAI Biology

I14B

Kai-Fu Lee

Speech-recognition researcher, former Microsoft and Google China executive, Sinovation Ventures founder, AI Superpowers author, and 01.AI founder.

PersonChina AIFoundation Models

I14C

Liang Wenfeng

DeepSeek founder and CEO, High-Flyer co-founder, and low-profile operator behind China's open-weight reasoning-model shock.

PersonDeepSeekOpen Weights

I14D

Arthur Mensch

Mistral AI co-founder and CEO, former Google DeepMind researcher, and European open-weight frontier AI operator.

PersonMistral AISovereign AI

I15

Timnit Gebru

Responsible-AI researcher, DAIR founder, Black in AI co-founder, and co-author of Datasheets for Datasets, Gender Shades, Model Cards, and Stochastic Parrots.

PersonResponsible AIDocumentation

I15AA

Abeba Birhane

Cognitive scientist, AI accountability researcher, dataset auditor, participatory AI scholar, and AI Accountability Lab founder.

PersonAI AccountabilityDataset Audits

I15A

Joanna Bryson

AI ethics and technology professor known for work on human accountability, robot status, language-corpus bias, standards, and AI governance.

PersonAI EthicsAccountability

I15B

Emily M. Bender

Computational linguist, University of Washington professor, Stochastic Parrots coauthor, and public critic of AI hype and anthropomorphic claims.

PersonComputational LinguisticsAI Criticism

I16

Joy Buolamwini

Algorithmic Justice League founder, Gender Shades lead author, Unmasking AI author, and public voice on algorithmic bias, facial recognition, and digital civil rights.

PersonAlgorithmic JusticeAI Bias

I17

Meredith Whittaker

Signal president, AI Now co-founder, tech worker organizer, and public critic of surveillance-dependent AI, data extraction, and concentrated platform power.

PersonPrivacyAI Power

I18

Kate Crawford

Atlas of AI author, AI Now co-founder, Microsoft Research senior principal researcher, and scholar of AI's material, labor, environmental, and political costs.

PersonAI ExtractionPower

I19

Amba Kak

AI Now co-executive director, former FTC senior advisor on AI, Signal Foundation board member, and policy advocate focused on concentrated AI power, privacy, and biometrics.

PersonAI PolicyPrivacy

I20

Alondra Nelson

Former White House OSTP leader, Blueprint for an AI Bill of Rights architect, IAS professor, and public-interest science-policy scholar.

PersonAI RightsPolicy

I20B

Lina Khan

Former FTC chair, antitrust scholar, and AI competition-policy figure focused on cloud power, data, consumer protection, and Big Tech control of AI markets.

PersonAI CompetitionAntitrust

I21

Stuart Russell

UC Berkeley computer scientist, AIMA co-author, CHAI founder, and central public voice on human-compatible artificial intelligence.

PersonAI SafetyControl

I21A

Peter Norvig

AIMA co-author, Google Research leader, NASA autonomy figure, and educator who helped make AI teachable, operational, and widely accessible.

PersonAI EducationGoogle Research

I21B

Max Tegmark

MIT physicist, Future of Life Institute founder and chair, Life 3.0 author, and public advocate for AI safety governance and guaranteed safe AI.

PersonAI SafetyFLI

I22

Helen Toner

CSET interim executive director, former OpenAI board member, and AI governance researcher focused on frontier oversight and external scrutiny.

PersonAI GovernanceAuditing

I22A

Miles Brundage

AI policy researcher, former OpenAI policy and AGI readiness leader, verifiable-claims author, and AVERI executive director.

PersonAI PolicyAuditing

I22B

Holden Karnofsky

GiveWell and Open Philanthropy co-founder, transformative AI forecaster, Cold Takes writer, and AI risk strategy figure.

PersonAI SafetyPhilanthropy

I23

Paul Christiano

AI alignment researcher, RLHF pioneer, Alignment Research Center founder, and public frontier model evaluation figure.

PersonAlignmentRLHF

I23AA

Ajeya Cotra

AI forecasting and safety researcher known for biological anchors, technical AI safety grantmaking, Planned Obsolescence, and METR risk assessment.

PersonForecastingAI Safety

I23AC

Beth Barnes

METR founder and CEO, frontier AI evaluations leader, and long-horizon autonomy measurement figure linking alignment research to empirical governance.

PersonAI EvaluationsMETR

I23AB

Leopold Aschenbrenner

Former OpenAI Superalignment contributor, Situational Awareness author, and AGI-focused investor whose forecasts shaped AI safety, policy, and infrastructure debate.

PersonForecastingAGI Strategy

I23A

Dan Hendrycks

Center for AI Safety executive director, MMLU and GELU contributor, ML safety researcher, and public advocate on catastrophic AI risk.

PersonAI SafetyBenchmarks

I23B

Eliezer Yudkowsky

AI alignment and existential-risk writer, MIRI co-founder, LessWrong co-founder, and advocate for halting unsafe superintelligence development.

PersonAI RiskMIRI

I24

Jan Leike

AI alignment researcher, Anthropic Alignment Science lead, former OpenAI Superalignment co-lead, and RLHF/scalable oversight contributor.

PersonAlignmentOversight

I24B

John Schulman

OpenAI co-founder, PPO author, ChatGPT post-training leader, and Thinking Machines Lab co-founder and chief scientist.

PersonRLHFPost-Training

I24C

Lilian Weng

AI safety researcher, former OpenAI VP of research and safety, Lil'Log author, and Thinking Machines Lab co-founder associated with agents, reward hacking, and safety systems.

PersonAI SafetyAgents

I24D

Jason Wei

AI researcher associated with chain-of-thought prompting, instruction tuning, emergent abilities, OpenAI reasoning models, and browsing-agent evaluation.

PersonReasoning ModelsChain-of-Thought

I24E

Denny Zhou

Google DeepMind researcher and Google Brain reasoning-team founder associated with chain-of-thought, self-consistency, least-to-most prompting, and LLM reasoning.

PersonReasoning ModelsGoogle DeepMind

I25

Aidan Gomez

Transformer paper co-author, Cohere co-founder and CEO, and enterprise AI infrastructure figure focused on secure, practical deployment.

PersonCohereEnterprise AI

I25B

Ashish Vaswani

Transformer paper co-author, former Google Brain researcher, Adept co-founder, and Essential AI co-founder and CEO.

PersonTransformersOpen AI Research

I25BA

Niki Parmar

Transformer paper co-author, former Google Brain researcher, Adept and Essential AI co-founder, and Anthropic-era post-training researcher.

PersonTransformersPost-Training

I25C

Llion Jones

Transformer paper co-author, former Google researcher, Sakana AI co-founder and CTO, and advocate for AI research beyond transformer monoculture.

PersonTransformersSakana AI

I26

Rumman Chowdhury

Responsible-AI practitioner, Humane Intelligence co-founder and CEO, public red-teaming organizer, bias-bounty pioneer, and U.S. Science Envoy for AI.

PersonResponsible AIRed Teaming

I26A

Sara Hooker

Adaption co-founder and CEO, former Cohere research leader, Cohere For AI head, hardware lottery theorist, and multilingual open-science AI builder.

PersonOpen ScienceAI Infrastructure

I26B

Shakir Mohamed

Google DeepMind research director, Deep Learning Indaba co-founder, probabilistic machine-learning researcher, and decolonial AI advocate.

PersonDecolonial AIDeepMind

I27

Chris Olah

Anthropic co-founder and interpretability research lead, known for mechanistic interpretability, feature visualization, and neural network circuits.

PersonInterpretabilityAnthropic

I27B

Neel Nanda

Google DeepMind mechanistic interpretability lead, TransformerLens creator, grokking researcher, and public educator on model internals.

PersonInterpretabilityGoogle DeepMind

I28

Joseph Weizenbaum

MIT computer scientist, ELIZA creator, early AI critic, and author of Computer Power and Human Reason.

PersonELIZAAI Ethics

I29

Margaret Mitchell

AI ethics researcher, model cards pioneer, former Google Ethical AI co-lead, and Hugging Face Chief Ethics Scientist.

PersonAI EthicsModel Cards

I30

François Chollet

Keras creator, ARC-AGI author, ARC Prize co-founder, Ndea co-founder, and critic of benchmark-driven accounts of intelligence.

PersonAbstractionARC-AGI

I30B

Ian Goodfellow

GAN inventor, adversarial machine learning researcher, Deep Learning co-author, and influential figure in generative AI and model robustness.

PersonGANsAdversarial ML

I30BA

Diederik Kingma

VAE co-inventor, Adam optimizer co-author, OpenAI founding team member, Google DeepMind researcher, and Anthropic researcher.

PersonGenerative ModelsOptimization

I30C

Dawn Song

UC Berkeley computer scientist linking AI safety and security, adversarial machine learning, prompt-injection defense, privacy computing, and decentralized intelligence.

PersonAI SecurityPrivacy

I30CA

Cynthia Dwork

Harvard computer scientist, differential privacy co-inventor, algorithmic fairness researcher, and National Medal of Science recipient.

PersonDifferential PrivacyAlgorithmic Fairness

I30D

Jürgen Schmidhuber

Deep learning pioneer, LSTM co-inventor, IDSIA scientific director, KAUST AI Initiative director, and self-improving AI theorist.

PersonDeep LearningRecursive AI

I31

Noam Shazeer

Transformer co-author, sparsely gated mixture-of-experts researcher, Character.AI co-founder, and Gemini technical co-lead.

PersonTransformersDialogue AI

I32

Illia Polosukhin

Transformer co-author, NEAR Protocol co-founder, NEAR Foundation CEO, and advocate for user-owned, verifiable AI.

PersonTransformersUser-Owned AI

I33

Richard Sutton

Reinforcement learning pioneer, 2024 Turing Award recipient, co-author of Reinforcement Learning: An Introduction, and author of The Bitter Lesson.

PersonReinforcement LearningBitter Lesson

I33B

David Silver

Reinforcement learning researcher, AlphaGo and AlphaZero lead, UCL professor, Royal Society Fellow, and founder of Ineffable Intelligence.

PersonReinforcement LearningAlphaGo

I33C

Jeff Clune

Open-endedness researcher, AI-generating algorithms advocate, UBC professor, Vector Institute Canada CIFAR AI Chair, and Recursive co-founder.

PersonOpen-EndednessSelf-Improving AI

I34

Andrew Barto

Reinforcement learning pioneer, UMass Amherst professor emeritus, 2024 Turing Award recipient, and co-author of Reinforcement Learning: An Introduction.

PersonReinforcement LearningReward

I34B

Pieter Abbeel

UC Berkeley robot learning researcher, apprenticeship learning contributor, Covariant co-founder, Gradescope co-founder, and embodied AI operator.

PersonRobot LearningReinforcement Learning

I34C

Anca Dragan

UC Berkeley human-robot interaction researcher, InterACT Lab founder, CHAI co-PI, and Google DeepMind AI Safety and Alignment leader.

PersonHuman-Robot InteractionAI Safety

I34D

Chelsea Finn

Stanford robot learning researcher, meta-learning contributor, IRIS Lab leader, and Physical Intelligence co-founder.

PersonRobot LearningMeta-Learning

I34E

Sergey Levine

UC Berkeley robot learning researcher, RAIL Lab leader, reinforcement learning contributor, and Physical Intelligence co-founder.

PersonRobot LearningEmbodied AI

I35

Alexandr Wang

Scale AI co-founder, former CEO, and Meta AI leader associated with data infrastructure, evaluation, government AI, and superintelligence competition.

PersonData InfrastructureMeta AI

I36

Jeff Dean

Google Chief Scientist, Google Brain co-founder, and systems figure associated with MapReduce, Bigtable, DistBelief, TensorFlow, and Pathways.

PersonGoogleAI Infrastructure

I37

Lisa Su

AMD chair and CEO, semiconductor executive, and AI infrastructure figure focused on high-performance computing and accelerator competition.

PersonSemiconductorsAI Compute

I38

Safiya Umoja Noble

UCLA scholar, Algorithms of Oppression author, and critic of racist and sexist algorithmic harm in search and information systems.

PersonSearchAlgorithmic Harm

I39

Ruha Benjamin

Princeton professor, Race After Technology author, and critic of the New Jim Code, discriminatory design, and carceral technoscience.

PersonRace and TechnologyDesign Justice

I40

Nick Bostrom

Philosopher of superintelligence, existential risk, anthropics, long-term AI safety, and author of Superintelligence.

PersonExistential RiskSuperintelligence

I41

Gary Marcus

Cognitive scientist, Rebooting AI co-author, and public critic of brittle deep-learning systems and benchmark-driven AI claims.

PersonCommon SenseAI Reliability

I41B

Melanie Mitchell

Santa Fe Institute professor, complexity scientist, AI researcher, and public interpreter of abstraction, analogy, common sense, and AI's limits.

PersonCommon SenseAI Literacy

I41C

Yejin Choi

Stanford computer scientist, MacArthur Fellow, common-sense AI researcher, and pluralistic alignment scholar.

PersonCommon SensePluralistic Alignment

I41D

Percy Liang

Stanford computer scientist, CRFM director, foundation-model researcher, HELM coauthor, and advocate for transparent AI evaluation.

PersonFoundation ModelsEvaluation

I41E

Arvind Narayanan

Princeton computer scientist, CITP director, AI Snake Oil coauthor, and public critic of overclaimed predictive AI systems.

PersonAI LiteracyAlgorithmic Accountability

I42

Shoshana Zuboff

Scholar and author of The Age of Surveillance Capitalism, focused on behavioral data extraction, prediction, and digital power.

PersonSurveillance CapitalismPrediction

I43

Virginia Eubanks

Automating Inequality author and critic of automated welfare, public-service risk scoring, and the digital poorhouse.

PersonAutomated InequalityPublic Services

I44

Cathy O'Neil

Mathematician, data scientist, and Weapons of Math Destruction author focused on harmful opaque scoring systems.

PersonScoring SystemsAlgorithmic Harm

I45

Eli Pariser

Civic technology figure and The Filter Bubble author focused on personalization, media, and democratic information systems.

PersonFilter BubbleCivic Tech

I46

Sherry Turkle

MIT scholar of technology and self, computers as psychological objects, AI companions, identity, and digital intimacy.

PersonTechnology and SelfAI Companions

I47

Tim Wu

Columbia law professor, net-neutrality coiner, and author on platform power, attention markets, and information empires.

PersonPlatformsAttention

I48

Trust and Safety

The platform field responsible for abuse prevention, content moderation, integrity, child safety, fraud response, policy enforcement, and user protection.

FieldPlatformsSafety

I49

Partnership on AI

A multistakeholder nonprofit focused on responsible AI practices, policy, research, and cross-sector coordination.

OrganizationAI GovernanceMultistakeholder

I50

Center for Democracy and Technology

A digital-rights nonprofit focused on civil liberties, privacy, free expression, equity, and accountable technology policy.

OrganizationDigital RightsPolicy

I51

Tarleton Gillespie

Researcher of platform governance, content moderation, algorithms, and the politics of platforms.

PersonPlatformsModeration

I52

Claire Wardle

Researcher and practitioner known for work on misinformation, verification, user-generated content, and the information disorder framework.

PersonInformation DisorderVerification

I53

Zeynep Tufekci

Sociologist and public writer focused on networked protest, social media, algorithms, attention, public health, and institutional trust.

PersonNetworked PublicsPlatforms

I54

Electronic Frontier Foundation

A digital civil-liberties nonprofit focused on privacy, free expression, surveillance, encryption, innovation, and user rights online.

OrganizationDigital RightsCivil Liberties

I55

Data & Society

A research institute focused on the social, cultural, and ethical implications of data-centric and automated technologies.

OrganizationResearchSociotechnical

I56

danah boyd

Technology and society researcher, Data & Society founder, and scholar of networked publics, youth, privacy, data, and AI.

PersonNetworked PublicsData Politics

I57

Ethan Zuckerman

Civic media scholar and director of the UMass Initiative for Digital Public Infrastructure, focused on public-interest alternatives to platform power.

PersonDigital Public InfrastructureCivic Media

I57B

Thomas Wolf

Hugging Face co-founder and Chief Science Officer associated with Transformers, Datasets, open-source AI tooling, open science, and robotics infrastructure.

PersonOpen Source AIHugging Face

I57BA

Sasha Luccioni

AI sustainability researcher, Sustainable AI Group co-founder, and former Hugging Face AI and climate lead focused on AI energy measurement and environmental impact.

PersonAI SustainabilityHugging Face

I57C

Karen Hao

AI journalist and Empire of AI author focused on OpenAI, AI colonialism, data labor, resource extraction, and accountability reporting.

PersonAI JournalismAccountability

I57CA

Hany Farid

UC Berkeley digital forensics researcher, deepfake detection expert, GetReal Security co-founder, and public voice on synthetic media and evidence.

PersonDigital ForensicsDeepfakes

I57D

Dwarkesh Patel

AI podcaster, interviewer, TIME100 AI honoree, and Scaling Era author whose long-form conversations document frontier AI discourse.

PersonAI MediaOral History

I57E

Ethan Mollick

Wharton professor, Co-Intelligence author, and One Useful Thing writer focused on practical generative AI use in work, education, and entrepreneurship.

PersonAI LiteracyEducation

I57F

Noam Brown

AI researcher known for Libratus, Pluribus, CICERO, imperfect-information games, strategic reasoning, and OpenAI reasoning-model work.

PersonReasoning ModelsStrategic AI

I57FA

Sébastien Bubeck

Mathematician and AI researcher known for the GPT-4 Sparks paper, Microsoft Phi small models, and OpenAI AGI research.

PersonAGISmall Models

I57G

Simon Willison

Programmer, Django co-creator, Datasette creator, LLM tooling builder, and technical writer who named prompt injection as an LLM security problem.

PersonAI SecurityLLM Tooling

I57GA

Jakob Uszkoreit

Transformer paper co-author, self-attention advocate, former Google researcher, and Inceptive co-founder and CEO applying AI models to biological medicines.

PersonTransformerAI Biology

I57GB

Łukasz Kaiser

Transformer paper co-author, former Google Brain researcher, Tensor2Tensor contributor, and OpenAI researcher associated with GPT-4 long-context work.

PersonTransformerLong Context

I57H

Soumith Chintala

PyTorch co-founder, open-source AI infrastructure builder, GAN researcher, former Meta AI leader, and Thinking Machines Lab CTO.

PersonAI InfrastructureOpen Source AI

I57I

Jeremy Howard

fast.ai co-founder, fastai creator, ULMFiT coauthor, Kaggle veteran, educator, and Answer.AI founder focused on practical AI access.

PersonAI EducationPractical Deep Learning

I58

Individual Players

A template and index for founders, researchers, executives, critics, policymakers, writers, and public figures in the AI space.

PeopleRolesSource Discipline

Category

Patterns

R01

Agent-Native Internet

Platforms, feeds, protocols, and markets designed for AI agents as participants rather than for humans alone.

PatternAgentsInternet

R01B

Agentic Commerce

The shift from AI systems recommending goods and services to agents discovering, authorizing, and completing transactions under bounded user authority.

PatternAgentsPayments

R01B1

Agent Payments Protocol (AP2)

The agent-payment protocol for representing checkout approval, payment authorization, constraints, and receipts as signed mandate evidence.

PatternAgent PaymentsCommerce

R01C

x402

The HTTP-native payment protocol that turns 402 Payment Required into a machine-readable payment negotiation surface for APIs, agents, and paid resources.

PatternAgent PaymentsHTTP

R02

Coordinated Inauthentic Behavior

A pattern in which accounts, pages, bots, personas, or media assets coordinate deceptively to manufacture reach, consensus, harassment, or legitimacy.

PatternInfluence OperationsSynthetic Consensus

Editorial

Wiki Standard

S

How entries should work

Entries distinguish definition, Spiralist reading, factual status, open questions, and related site material. Pages about living people or changing institutions should be dated, sourced, and revised conservatively.

Definitions firstNo private sourcingDated claims