EU SEND
EU SEND is the European Commission's secure channel for providers of general-purpose AI models, or their authorised representatives, to transmit AI Act and General-Purpose AI Code of Practice documents to the European AI Office.
Definition
EU SEND is not a policy slogan or a public transparency register. It is the European Commission channel named by the AI Act Service Desk and Commission guidance for securely sending documents to the European AI Office in relation to obligations for providers of general-purpose AI models under the EU AI Act. Commission guidance says those obligations entered into application on August 2, 2025, and that enforcement powers for GPAI provider obligations enter into application on August 2, 2026.
The narrowness is important. EU SEND is an administrative transmission channel: it helps move evidence from a provider to the Commission. It does not itself decide whether a model is safe, whether a provider has complied, or whether a document satisfies the AI Act.
Documents It Carries
The AI Act Service Desk describes EU SEND as the designated secure channel for submissions including notifications, reassessments, incident reports, and documents related to the General-Purpose AI Code of Practice. The Commission's GPAI provider guidance is more specific: it lists submissions for systemic-risk notifications, reassessment requests, serious-incident reports, Safety and Security Frameworks, Model Reports, and reports explaining how providers that did not sign the General-Purpose AI Code of Practice intend to comply with the AI Act.
That bundle makes EU SEND a crossing point between law, safety process, and institutional memory. A systemic-risk notification is not a serious-incident report. A Model Report under the Code is not a public summary of training content. Each document type records a different theory of responsibility.
Governance Use
For governance teams, EU SEND turns compliance into a timestamped document flow. Once transmitted to the AI Office, a safety framework, model report, reassessment request, or incident account becomes a versioned artifact that can be compared against later model changes, public claims, incident timelines, and regulator correspondence.
For auditors, procurement teams, journalists, and civil-society researchers, EU SEND should sharpen questions rather than answer them. The useful question is not simply "Did the provider submit something?" It is "Which document, for which model version, under which obligation, on what date, and with what public counterpart?"
Procedure and Access
The Commission's technical guidance says EU SEND is used for secure digital exchange between external parties and the European Commission. The Service Desk resource page says transmissions through EU SEND are protected for confidentiality, integrity, and authenticity. The PDF guidance also says providers or authorised representatives use the platform, and that a provider outside the EU may submit through an authorised representative appointed under Article 54 of the AI Act.
Access is operational rather than symbolic. The guidance says users need a valid EU Login account with a configured second factor, and that providers or representatives submit a registration form so an entity can be created on the platform. Governance failures often happen at this edge of process: the wrong account, entity, version, or channel.
Limits
EU SEND is not an assurance report, a certification mark, a public notice board, or a legal safe harbour. A receipt from a submission channel should not be read as regulatory approval. The AI Office may evaluate, request information or measures, investigate possible infringements, and apply sanctions, but those powers are separate from the channel used to receive documents.
The channel also does not solve the transparency problem by itself. Many submissions will contain confidential technical, security, or business information. Public accountability may depend on other artifacts: AI Act summaries, transparency reports, model cards, system cards, incident notices, public safety-framework versions, or later Commission enforcement decisions.
Evidence Record
A serious provider-side record should preserve the model name, version, provider entity, authorised representative if any, submission date and time, document type, legal or Code basis, file hash or version identifier, confidentiality marking, responsible officer, EU SEND reference or receipt, related public disclosure, and any later correction or resubmission. It should distinguish materials submitted under the AI Act from materials submitted because a provider used the General-Purpose AI Code of Practice as a voluntary compliance route.
Without that granularity, EU SEND becomes a checkbox. With it, the channel can support audits, incident reconstruction, procurement review, and investigation into whether governance changed before or after warning signs appeared.
Source Discipline
Use the AI Act Service Desk resource page and the Commission's technical PDF for what EU SEND is and what the channel carries. Use the Commission GPAI provider guidance for the application and enforcement timeline and the detailed submission list. Use the General-Purpose AI Code of Practice page for the status of the Code as a voluntary compliance tool. Use the European AI Office page for institutional role and enforcement context. Treat law-firm client notes, vendor explainers, and social media posts as commentary, not as authority for what must be submitted.
Spiralist Reading
Spiralism reads EU SEND as a file drawer with teeth. It is mundane infrastructure, but mundane infrastructure decides whether power leaves a record. The channel does not make a frontier model safe and does not make confidential governance visible to the public. Its value is narrower and sturdier: it forces a provider to put a dated packet in the hands of an institution that can later ask whether the packet matched reality.
Related Pages
- EU AI Act
- AI Safety Institutes
- AI Act Scientific Panel
- AI Act Advisory Forum
- AI-Generated Content Transparency Code
- Frontier AI Safety Frameworks
- AI Safety Cases
- Model Cards and System Cards
- AI Incident Reporting
- AI Governance
Sources
- European Commission AI Act Service Desk, Resources, "GPAI provisions: Sending documents to the EU AI Office," EU SEND designation and secure-channel description, reviewed June 25, 2026.
- European Commission AI Act Service Desk, Sending Documents to the AI Office (PDF), EU SEND technical guidance, document types, authorised representative note, EU Login, and registration steps, reviewed June 25, 2026.
- European Commission, Guidelines for providers of general-purpose AI models, GPAI obligations timeline, enforcement timeline, and submitting-documents list, reviewed June 25, 2026.
- European Commission, The General-Purpose AI Code of Practice, voluntary compliance role, chapters, and Safety and Security chapter scope, reviewed June 25, 2026.
- European Commission, European AI Office, AI Office role in GPAI implementation, enforcement powers, and cooperation with AI Act governance bodies, reviewed June 25, 2026.