AI Office

in English: European AI Office

basicspreading

The European Commission's expert body, the European AI Office, which coordinates implementation of the EU AI Act and directly supervises developers of general-purpose AI models.

In plain terms

The AI Office is the European Union's 'AI supervision'. It operates like a market surveillance authority: it does not manufacture the product nor approve it in advance, but holds developers accountable for complying with common rules. Here, the product is the large language or image-generating model on which chatbots, assistants, and AI agents are built.

The office is not an independent institution but part of the European Commission, located in Brussels. Lawyers, engineers, economists, and AI safety researchers work there. Its job is to ensure the AI Act works in practice: it writes guidelines, evaluates models, and coordinates the work of national authorities.

It differs from a national authority in that it is the sole supervisor of general-purpose AI model developers across the entire EU. Thus, a large model manufacturer does not have to coordinate with 27 authorities in 27 countries. However, specific high-risk applications (such as a recruitment or credit scoring system) are typically monitored by member state authorities.

How it works

Interpreting rules: it issues guidelines and templates on what the regulation's requirements mean in practice, for example, what a model developer must document.
It maintains a code of practice: voluntary commitments developed jointly with developers, researchers, and civil society organizations, which companies can sign to demonstrate compliance more easily.
It examines models: it can request documentation and information from developers, evaluate the most powerful so-called systemic-risk models, and receive reports on serious incidents.
It acts if needed: it can prescribe risk-reduction measures and, in extreme cases, request model withdrawal. The Commission may impose a fine based on the agency's work, up to 3 percent of the company's global annual turnover or 15 million euros (whichever is higher).
Coordinates: works with member state authorities, the AI board grouping member states, and an independent scientific expert panel, and participates in the EU's international AI coordination efforts.

An everyday example

You open a chatbot and read on the provider's website that the model's developer has signed the EU Code of Practice and published a summary of the training data. Behind these stand the AI Office's templates and expectations. Part of this rulebook is also that the system must signal that you are talking to a machine.

Who is affected

developerIf you are developing a general-purpose model or building a product on top of one, the authority's guidelines and templates determine what you must document and hand over. → Code

entrepreneur, executiveYou can learn from the authority's guidelines what obligations fall on your company if you implement or sell AI systems in the EU.

lawyer, citizenThis body shapes the practical interpretation of the AI Act, meaning how your rights are enforced against artificial intelligence. → Law

everyoneThis EU-level agency demands accountability for the security and transparency of the models behind chatbots and assistants used daily.

A common misconception

MythThe AI Office approves every AI application before it can be launched on the EU market.

In realityThere is no general prior authorization, and the authority does not examine every application individually. It primarily supervises developers of general-purpose models, while the inspection of most specific AI systems is the responsibility of national authorities.

Limits and risks

For an individual complaint, you typically go not to the Office but to the national authority, and in data protection matters to the data protection authority.
The Code of Practice is voluntary: a company having signed it does not mean its product is flawless or risk-free.
The authority's staff and resources are modest compared to those of the supervised global technology companies, so the depth of oversight may be limited.
Rules come into force gradually and deadlines may change, so always check the latest official information.

Similar, but not the same

MI-rendelet (AI Act) – The AI Act is the law itself, while the AI Office is the organization that supports and partially enforces its implementation.
National market surveillance authority – National authorities monitor AI systems used in their respective countries, while the EU AI Office oversees general-purpose models at the Union level.
European Artificial Intelligence Board (AI Board) – The body is an advisory and coordination forum for representatives of member states, while the office is the Commission's permanent apparatus carrying out daily professional work.

Related concepts

Broader

MI-rendelet (AI Act), AI regulation, AI governance, European Commission

Narrower

Code of Practice for General-Purpose AI, Systemic risk, Scientific expert panel

On the edge

General-purpose AI model (GPAI), High-risk AI system, European Artificial Intelligence Board, Regulatory sandbox, AI safety institute

A brief history

The European Commission established the office by a decision in January 2024, and it began operating in February 2024 within the Commission's directorate-general responsible for digital affairs. Its tasks are laid down in the AI Act, which entered into force on 1 August 2024. Obligations for general-purpose models apply from 2 August 2025; the code of practice was published shortly before, in July 2025.

Where things stand

The office is operational: it issues guidelines, maintains the code of practice, and since 2 August 2026 the European Commission's enforcement powers against developers of general-purpose models, including fines, have been in effect. Actual supervisory practice is still taking shape. Deadlines for high-risk systems were pushed back by the 2026 EU simplification package (digital omnibus), so for these, always check the latest official dates.

Status: 2026 November

In the news

Appears in 10 Agent Age stories; the curve shows weekly occurrence in the news magazine.

UAE and Saudi Arabia launch vibe coding programmes and guidanceOct 29, 2026 · Code
WHO European Region adopts 2026–2033 health AI roadmap; Luxembourg to host officeOct 26, 2026 · Health
EU AI Office opens first probes into hiring and credit-scoring AIOct 5, 2026 · Work & HR
GE HealthCare hires Medtronic's AI leader as chief AI officerSep 29, 2026 · Health
EU AI rules follow HR content across borders as AI Office inspections target hiringSep 28, 2026 · Work & HR
Reports: EU AI Act compliance inspections to begin, with HR in scopeSep 14, 2026 · Work & HR
AI Act enforcement begins as the AI Office starts requesting information from companiesAug 31, 2026 · Jog
AI Act enforcement and transparency obligations apply from 2 August 2026Aug 17, 2026 · Jog

Also known as

AI Office · EU AI Office · European AI Office · European Artificial Intelligence Office · MI-hivatal · European AI Office · Artificial Intelligence Office · EU AI Office

Sources

European Commission – European AI Office

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.