The European AI Office is a European Commission body, established in 2024 within DG CONNECT, that supervises, enforces, and coordinates implementation of the EU AI Act's rules for general-purpose AI (GPAI) models, including systemic-risk models. It acts as the central EU-level coordinator for AI governance and is the direct supervisor of GPAI model providers under the Act. As of late 2025 it had more than 125 staff, with further recruitment planned.
| Type | European Commission body (within DG CONNECT) |
| Established | 2024 |
| Role | Supervise, enforce, and coordinate implementation of the EU AI Act's rules for general-purpose AI (GPAI) models, including systemic-risk models |
| Staff | 125+ (late 2025), with further recruitment planned |
Organisation
As of late 2025 the AI Office was organised into six operational units plus two advisory positions (Source: EU AI Office — GPAI Provider Guidelines and Enforcement Framework). The units are Excellence in AI and Robotics; Regulation and Compliance; AI Safety; AI Innovation and Policy Coordination; AI for Societal Good; and AI in Health and Life Science. The two advisory positions are a Lead Scientific Advisor and an Advisor for International Affairs. Staffing includes technology specialists, administrative assistants, lawyers, policy specialists, and economists.
Functions and activities
The AI Office supervises and enforces the EU AI Act's GPAI rules and serves as the central EU-level coordinator for AI governance.
For general-purpose AI models under EU AI Act Articles 51–56, the AI Office is the sole regulator for providers, a centralisation at Commission level rather than a distribution to member states, attributed to the technical capacity that GPAI supervision requires. It chairs the Signatory Taskforce for the GPAI Code of Practice and led the multi-stakeholder process that drafted the Code. It issued the Guidelines on the scope of obligations for providers of general-purpose AI models under the AI Act (last updated 26 March 2026), the Commission's authoritative interpretation of who counts as a GPAI provider (Source: EU AI Office — GPAI Provider Guidelines and Enforcement Framework). It runs the EU SEND platform for standardised submission of GPAI provider documentation.
The Commission's EU KIDS Act proposal of 17 September 2026 would extend the Office's remit to child protection. Obligations on providers of AI companions and general conversational chatbots would be supervised under the AI Act's enforcement chapter, with the Commission supervising the systems under its exclusive competence under Article 75(1) of the AI Act, the AI Office named as the complaint venue for those systems, and an expedited procedure aiming at preliminary findings within [30] working days and a final decision within 90 working days. The Commission's explainer describes this as direct Commission supervision of the most widely used chatbots (Source: digital-strategy.ec.europa.eu); the proposal's financial statement describes the Commission as responsible for AI companions and chatbots built on general-purpose AI models or integrated into very large online platforms or search engines. The proposal's financial statement provides for 10 non-permanent full-time equivalents to be redeployed within the Commission to reinforce the AI Office, and its staffing tables list a further 5 full-time equivalents of external staff under an AI Office line (Proposal for an EU KIDS Act, COM(2026) 681 — Source Summary). Non-compliance by companion and chatbot providers could draw fines under AI Act Article 99 of up to 6% of worldwide annual turnover where the provider acted intentionally or negligently, and the Commission's AI Act powers under Articles 75a to 75d would extend to the new obligations (Proposal for an EU KIDS Act, COM(2026) 681 — Source Summary).
On coordination and advisory functions, the AI Office supports the European AI Board (member-state representatives) and provides the secretariat for both the Scientific Panel and the Advisory Forum. It represents the EU in international AI governance fora. It also leads ecosystem-development efforts, including the AI Continent Action Plan (April 2025), the Apply AI Strategy (October 2025), and GenAI4EU. On 19 July 2026 the AI Office published expert findings for the European Frontier AI Initiative identifying compute and energy as the urgent two-year priorities for EU competitiveness, alongside growth-stage capital and legal certainty on training data (Source: digital-strategy.ec.europa.eu).
Enforcement status and member-state readiness
GPAI obligations under the AI Act have applied since 2 August 2025, and the AI Office is receiving systemic-risk notifications, SSF reports, and training-data summaries. Commission enforcement powers, including fines, activate on 2 August 2026, and the AI Office is the instrument that will exercise those powers. The Digital Simplification Package proposed in November 2025 included amendments that would further centralise GPAI oversight at the Commission (Source: EU AI Office — GPAI Provider Guidelines and Enforcement Framework).
The broader EU enforcement architecture is behind schedule relative to the AI Office's own readiness. Member states were required to designate national competent authorities and single points of contact by 2 August 2025. As of March 2026, seven months past that deadline, only 8 of 27 Member States had designated a single point of contact (Source: EU AI Office — GPAI Provider Guidelines and Enforcement Framework). This shortfall affects enforcement of high-risk AI systems, which relies on member-state market-surveillance authorities, rather than GPAI supervision, which the Commission handles directly. The November 2025 Digital Simplification Package proposal is partly responsive to this gap.
The Office made first use of those powers during the week of August 24, 2026, ordering information from developers of the most advanced models on cybersecurity, safety and copyright compliance. Executive Vice-President Henna Virkkunen described the requests in an interview published August 26, 2026 without naming recipients, saying "all the providers are not covered with this request for information because, with some of them, we have already had very close dialogue during the last months." The security requests cover mitigations "to prevent humans or AI from stealing the model," external-evaluator access, follow-up on outside safety recommendations, and post-release monitoring; a separate exercise asked more than 30 companies about EU copyright compliance, directed at developers that had not published the required training-data summary or had not answered informal inquiries (Source: euractiv.com). As of an August 28, 2026 account the Office had not pursued any covered company for misconduct (Source: axios.com).
Relationship to the AI Act timeline
The AI Act entered into force on 1 August 2024. Its prohibitions (Article 5) applied from 2 February 2025. GPAI obligations (Article 53) applied from 2 August 2025, the same date the GPAI Code of Practice was endorsed (1 August 2025) and the deadline for member-state authority designation. Commission GPAI enforcement powers, including fines of up to €15M or 3% of global turnover, activate on 2 August 2026, alongside full application of most high-risk provisions. By 2 August 2027, pre-2025 GPAI models must be brought into compliance and high-risk system obligations fully apply.
Institutional partners
The AI Office works with the European Artificial Intelligence Board (member-state representatives, implementation coordination); the Scientific Panel (independent experts, systemic-risk advisory); and the Advisory Forum (industry, SMEs, civil society, academia). It maintains a partnership with the European Centre for Algorithmic Transparency and engages industry through the voluntary AI Pact and Apply AI Alliance platforms.
Relationships
- supports: EU AI Act (Regulation 2024/1689), EU General-Purpose AI Code of Practice (Final Version, 2025)
- related: EU KIDS Act (proposed Regulation on the protection of minors online) — proposed supervisor of AI companion and chatbot obligations toward minors.
- related: EU AI Office — GPAI Provider Guidelines and Enforcement Framework, Frontier Compliance Framework (February 2026), Anthropic's Responsible Scaling Policy (Version 3.1), OpenAI Preparedness Framework V.2