Supervisory authority AI: who enforces the EU AI Act
Who the supervisory authority for AI is under the EU AI Act: designated national authorities, the AI Office's powers, pending designations, and which authority
In this article6

The authorities actually designated so far
Member States had to designate and notify their national competent authorities, meaning notifying authorities and market surveillance authorities, by 2 August 2025 under Article 70 of the AI Act (Regulation (EU) 2024/1689). Where a state missed that date, the Commission may open a formal infringement procedure.[1] The Commission's published list of Single Points of Contact was last updated on 7 September 2026 and is updated continuously.[1]
- Ireland's entry, the AI Office of Ireland, appears without an asterisk, meaning the designation decision has been adopted.[1]
- Entries marked with an asterisk mean the national designation decision is still pending final adoption. Those include Germany's Bundesnetzagentur, France's DGCCRF and Spain's Agencia Española de Supervisión de Inteligencia Artificial, as well as entries for Luxembourg, Cyprus, the Czech Republic, the Netherlands, Estonia, Portugal, Romania, Slovenia and Sweden.[1]
- Italy's Single Point of Contact is the Agenzia per la cybersicurezza nazionale, the National Cybersecurity Agency, listed without an asterisk as of 7 September 2026.[1]
- The European Data Protection Supervisor (EDPS) is the designated market surveillance authority for EU institutions, bodies and agencies.[2]
| Body | Stated role on its own page | AI Act supervisory role |
|---|---|---|
| BfDI | Federal Commissioner for Data Protection and Freedom of Information (Germany) | – |
| CEN-CENELEC | European standardisation organisation | – |
| CNIL | French data protection authority; publishes AI guidance and sanctions | – |
| Data Protection Commission | Irish supervisory authority for the GDPR | – |
| European Commission | Hosts the AI Office | Supervises and enforces the rules for general-purpose AI models |
| European Data Protection Board | Issues binding GDPR guidance; put GDPR guidelines on when to impose fines out for consultation in September 2026 | – |
| European Data Protection Supervisor | Supervises EU institutions' data processing | Market surveillance authority and notified body for EU institutions, bodies and agencies |
| European Parliament | – | – |
| Garante per la protezione dei dati personali | Italian data protection authority; imposed a GDPR fine on Character.AI in July 2026 | – |
| ISO | – | – |
Each cell was read from that body's own public page, retrieved on 23 September 2026. A dash means the page lists no value for that column. The bodies without an AI Act supervisory role remain relevant to readers through their GDPR or standardisation work, not as AI Act supervisors. This article was last verified on 23 September 2026; the next scheduled review is 23 March 2027.
Mandate, scope and powers under the Act
Market surveillance authorities investigate and enforce compliance with the rules for AI systems, including the prohibitions and the high-risk rules. Their powers include intervening when systems pose risks or do not comply, conducting remote monitoring, accessing providers' documentation, data sets and source code, proposing joint investigations with the Commission, requesting corrective measures and imposing penalties. They must accept complaints from any natural or legal person who has grounds to consider an infringement.[1]
- The AI Office, established within the European Commission, supervises and enforces the obligations for providers of general-purpose AI models. Its powers are to request information, conduct model evaluations, request measures from providers including risk mitigations and recalling a model from the market, and to impose fines of up to 3% of global annual turnover or 15 million euros, whichever is higher.[8]
- The AI Office sits within a governance structure of three advisory bodies: the European Artificial Intelligence Board, the Scientific Panel and the Advisory Forum.[5]
- Fundamental rights protection authorities, which Member States had to identify and publicly list by 2 November 2024, hold the right to be informed by market surveillance authorities, for example about serious incidents, and the right to request information, documentation and cooperation from them.[5]
How national authorities relate to the AI Office
The Commission and the AI Office work closely with the national competent authorities, who supervise the implementation and application of the AI Act at national level. Market surveillance authorities from different Member States come together under the framework of the European Artificial Intelligence Board to coordinate enforcement and exchange expertise, making use of existing EU tools for information exchange and cooperation. Where both levels have competence, the split follows the subject: the AI Office for general-purpose AI models, the national authority for AI systems placed on the market or used in that Member State.[5][1]
Where designations are still pending
The distinction between what applies and what is pending matters. An asterisked entry on the Commission list is a designation decision that has not yet been finally adopted; it is not an authority in force.[1] Penalty levels for prohibited practices and the enforcement architecture behind them follow the Act's separate fine tiers, set out in EU AI Act fines.[9]
Which authority applies to you
The answer follows from what your organisation places on the market or uses, and the application dates attached to each rule set.[10]
- Provider of a general-purpose AI model: the AI Office within the European Commission. The GPAI and governance rules apply since 2 August 2025.[8]
- Provider or deployer of an AI system in a Member State: that state's market surveillance authority or Single Point of Contact, checked against the Commission list for pending status.[1]
- EU institution, body or agency: the EDPS, which acts as market surveillance authority and notified body for EU institutions.[2]
The application dates that trigger obligations: prohibitions and AI literacy from 2 February 2025; GPAI and governance rules from 2 August 2025; most remaining provisions from 2 August 2026, except the high-risk rules; high-risk rules for Annex III use cases from 2 December 2027; and for Annex I products from 2 August 2028.[10] Providers of high-risk systems also face conformity assessment before placing a system on the market, and those obligations follow the same dates.[9]
The check to run before relying on any of this in a compliance plan: confirm the named authority and its status on the Commission's Single Points of Contact list, and note the list's last-update date, before treating a designation as in force.[1] Where a penalty question drives the plan, the tiered fine structure determines the exposure; the Penalty tiers are broken down separately.
One methodological note for readers tracing any figure in this article: the AI Act text itself is available in the Official Journal version of 12 July 2024,[9] and the Commission's AI Act pages, from its regulatory framework overview[11] to its main portal,[12] carry stated last-update dates. The CNIL's AI guidance and sanctions are published on its own site, in French and English.[13]
Sources
Rankings and scores on Optinest cannot be bought — no commercial relationship changes how a provider is assessed. How we score
Related reading
All Regulation →AI Act enforcement: who fines whom, how much and from when
Who enforces the EU AI Act, the four fine tiers from EUR 7.5 million to EUR 35 million, and which dates apply after Regulation (EU) 2026/1744.

When the AI Act requires a notified body
When the EU AI Act requires a notified body, when internal control suffices, how bodies are designated and the timeline set by the AI Omnibus.

What an AI vendor's DPA must put in writing
What a DPA for AI vendors must contain under GDPR Article 28, how ten European providers differ, and which claims a contract cannot back.
