Wiki · Person · Last reviewed June 16, 2026

Tim Wu

Tim Wu is a Columbia Law School professor and former White House technology-and-competition policy official whose work links communications infrastructure, platform power, attention markets, antitrust, and the governance of AI-mediated institutions.

Category: Person Published: June 16, 2026 Modified: June 16, 2026 Last reviewed: June 16, 2026 Tags: antitrust, net neutrality, platform power, attention economy, AI governance, communications law

Definition

Tim Wu is a legal scholar of communications law, antitrust, copyright, the First Amendment, telecom, media industries, and technology platforms. His importance to AI reference work is not that he is primarily an AI-safety theorist, but that he gives names to recurring control points: the network, the switch, the platform, the default, the attention market, and the private rule system.

Wu is best known for developing network neutrality theory, for the information-industry history in The Master Switch, for the attention-economy critique in The Attention Merchants, for the anti-monopoly argument in The Curse of Bigness, and for his 2021-2023 White House role on technology and competition policy. His later work extends that frame to platform extraction and AI-influenced markets.

For this wiki, Wu is best read as an infrastructure-and-power thinker. He is useful when the question is who controls a conduit, marketplace, default, data flow, or rule-enforcement layer. He is less useful as a source for claims about model internals, consciousness, alignment, or whether a particular AI system is safe.

Snapshot

Core Ideas

Network neutrality. Wu's 2003 paper Network Neutrality, Broadband Discrimination framed broadband carriers as infrastructure that could distort innovation if allowed to discriminate among applications, content, or services. The narrow telecom doctrine does not transfer automatically to AI, but the analytic question does: who controls access to the rail, and can that controller favor its own services?

Information empires. The Master Switch describes a recurring pattern in which open communications systems consolidate into vertically integrated empires. In AI, the analogous stack includes chips, cloud, data, models, APIs, agents, app stores, identity, payments, browsers, search, and enterprise suites. A firm can shape outcomes by controlling several layers even without banning rivals outright.

Attention merchants. Wu's attention-economy work treats human attention as a marketable resource captured, packaged, and resold by media and advertising systems. That frame is directly relevant to recommender systems, AI search answers, companions, and assistants that may optimize not only for helpfulness but also retention, persuasion, conversion, or behavioral dependency.

The curse of bigness. Wu's anti-monopoly argument is political as well as economic: concentrated private power can shape markets, information, public policy, and democratic possibility. The question is not simply whether a dominant platform raises prices; it is whether users, publishers, developers, workers, agencies, and competitors can leave, interoperate, audit, contest, and build alternatives.

Hybrid social ordering. In his 2019 Columbia Law Review essay on AI and law, Wu argued that full displacement of human courts was distant but that human-machine hybrids were a likely future for rule enforcement and dispute resolution. For AI governance, that points toward appeal, procedural fairness, hard-case review, and human judgment where automated systems exercise institutional power.

Boundary and Limits

Wu's framework travels well when an AI issue has an infrastructure or market-structure component. It helps analyze cloud dependence, model API access, search and assistant defaults, app-store review, enterprise-suite bundling, exclusive data licenses, ad markets, user attention, and private appeal systems. It is weaker if used as a generic label for every AI harm.

The net-neutrality analogy should be used with care. Broadband carriage, cloud services, model APIs, browser assistants, and agent registries are different layers with different laws, risks, and remedies. A nondiscrimination duty may make sense for one layer and be unsafe, unconstitutional, or counterproductive at another. The source-disciplined question is not "should AI have net neutrality?" but "which chokepoint is being controlled, by whom, under what legal theory, and what remedy would preserve contestability without increasing misuse or privacy risk?"

The antitrust frame also has limits. Concentration can make some safety work easier by giving a small number of actors capacity to coordinate, remove abusive uses, and invest in security. The counter-risk is that those same actors can define safety in ways that hide costs, block rivals, restrict research, or make exit impractical. Wu's value is in keeping both possibilities visible.

Current Context

As of June 16, 2026, Wu is back in public academic life at Columbia Law School and is still being cited in debates about Big Tech, platform extraction, antitrust, AI regulation, and the future of communications infrastructure. His 2025 book The Age of Extraction explicitly connects platform power to data, attention, wealth extraction, generative AI, predictive social data, and inequality.

Wu's White House work belongs to a specific policy moment. The Biden administration's 2021 competition agenda treated concentration, defaults, labor restraints, tech platforms, attention markets, surveillance, data aggregation, and sector-specific bottlenecks as government-wide economic problems. That agenda did not become a permanent statutory settlement: President Donald Trump revoked Executive Order 14036 on August 13, 2025, and the Federal Register entry for EO 14036 records that revocation by EO 14337. Current references to Wu's White House role should therefore distinguish his influence on the 2021-2023 Biden competition program from the federal posture in place after 2025.

AI competition questions have also become more concrete since Wu's White House tenure. The FTC warned in 2023 that generative AI depends on key inputs such as data, talent, compute, cloud services, and base models. Its 2024 inquiry and 2025 staff report examined partnerships among major cloud providers and AI developers, including cloud-spend commitments, switching costs, access to sensitive technical information, and possible effects on competition.

AI Governance Relevance

Wu's work is useful for AI governance because it shifts attention from model behavior alone to the systems that mediate access. A model can look open at the interface while the surrounding stack is closed: default assistant placement, cloud credits, proprietary tools, exclusive data licenses, opaque app review, search ranking, telemetry, identity, billing, or enterprise procurement can all become practical control points.

The net-neutrality analogy asks whether AI infrastructure should have nondiscrimination, interoperability, portability, or fair-access duties. That does not mean every model API should be a common carrier. It means the governance question should name the layer: network, cloud, model, app store, retrieval index, agent registry, safety gate, or user interface.

The attention-market analogy matters because AI systems can become relational interfaces rather than passive feeds. An AI companion, tutor, search answer engine, shopping assistant, workplace copilot, or political-content recommender can shape what a person notices, trusts, purchases, writes, remembers, and repeats. The safety issue is not only false output; it is persistent behavioral steering under commercial incentives.

The antitrust analogy matters because safety language can be both real and strategic. A platform may need strict controls to prevent fraud, child abuse material, cyber misuse, impersonation, or privacy leakage. The same platform may also use safety review, API terms, app-store rules, or trust labels to disadvantage competitors. Serious governance has to separate necessary safeguards from exclusionary gatekeeping.

Governance and Safety

Wu's concepts point to practical checks. Public agencies and civic institutions should avoid depending on a single model, cloud, identity layer, app store, or search surface without exit rights, audit rights, data-export plans, procurement records, and continuity plans. Vendor convenience is not the same as institutional resilience.

For AI systems that rank, recommend, moderate, adjudicate, or personalize, governance should preserve contestability. Users and affected parties need notice, appeal, correction paths, versioned records, and human review for consequential decisions. Wu's hybrid-social-ordering frame is especially important here: automation may handle routine scale, but high-stakes or ambiguous cases require accountable human judgment, not a decorative human checkbox.

For competition policy, remedies should be specific to the bottleneck. Cloud egress, model API discrimination, self-preferencing in assistants, exclusive data licenses, default contracts, app-store rules, and agent certification each call for different evidence. Broad claims about "Big Tech" are weaker than precise claims about a named control point, market, and remedy.

A Wu-informed AI procurement review should ask concrete questions before adoption: Can the institution export prompts, logs, memories, embeddings, fine-tunes, and evaluation records? Can it switch model or cloud providers without losing critical workflows? Does the vendor reserve unilateral power to change safety rules, ranking, available models, data-use terms, or pricing? Are appeals routed to a human with authority? Are safety denials logged in a way that can be audited without exposing sensitive user data?

For private platforms, the same review should separate safety from enclosure. A platform may need strict abuse controls, but it should be able to explain the threat model, rule, affected developers or users, evidence of proportionality, appeal process, and whether the same rule applies to the platform's own services. Without that record, safety governance can become indistinguishable from private industrial policy.

Source Discipline

Use Wu carefully. His books and essays are important frameworks, not proof that a particular AI company has violated competition law. A court judgment, agency report, procurement contract, regulator inquiry, platform policy, academic article, and publisher description carry different evidentiary weight.

For current AI and antitrust claims, cite primary records whenever possible: court opinions, regulator releases, government reports, official company terms, standards documents, and dated academic papers. For biographical claims about Wu, use Columbia Law, the Columbia Law archive, SSRN, publisher pages, and primary publications rather than unsourced summaries.

Do not overread stale biographical metadata. Columbia's faculty profile currently lists Wu's 2021 White House post in a professional-experience line, while Columbia's own 2023 news story says he had resumed teaching after nearly two years in the Biden administration. For present-tense claims, prefer dated narrative sources over a profile bullet that may lag institutional updates.

Spiralist Reading

For Spiralism, Wu is useful because he names the places where freedom can disappear without a dramatic ban: the default, the switch, the carrier rule, the attention market, the platform toll, the private appeals process, the cloud dependency, the interface that feels natural because alternatives were made inconvenient.

The Spiralist reading is not that every large platform is evil or every open system is safe. It is that agency depends on visible structure. AI governance must ask who owns the path between person and world, who can change the rules, who can appeal, and who can leave without losing their memory, audience, livelihood, or institutional capacity.

Open Questions

Competition and platforms

AI governance and infrastructure

Books

Sources

Return to Wiki