Enforcement of the AI Act Starts Today
Non-compliance may result in fines. Here's what you need to know | Edition #311
As expected, enforcement of the EU AI Act starts today, August 2.
Non-compliance may result in fines and can also affect companies based outside the EU that, for example, offer AI models or systems to customers in the EU (Article 2 of the AI Act provides more information).
Here is how enforcement will work in practice, and other information that everyone should know:
1. The AI Office will enforce the rules for providers of general-purpose AI models, including the rules for providers of models posing systemic risk.
2. Enforcement of transparency obligations and prohibited AI practices also starts today (with two exceptions, see item 4 below). Responsibility for enforcing these rules is shared across three different bodies:
The AI Office will enforce the rules for AI systems offered by the same provider as the underlying general-purpose AI model.
National competent authorities (each of the 27 Member States has its own authorities) will enforce the rules for other AI systems.
The European Data Protection Supervisor will enforce the rules for AI systems used by EU institutions, bodies, and agencies.
The AI Office has recently appointed Prof. Alessandro Abate as the Lead Scientific Adviser. He will support the Office’s scientific work on general-purpose AI models.
3. The EU launched tools to help people support the enforcement of the AI Act:
People can use this Complaint Tool to report alleged infringements of the AI Act by providers of AI systems supervised by the AI Office.
People working with providers of AI systems or general-purpose AI models can use the Whistleblower Tool to confidentially report possible violations of the AI Act.
There is also a channel available to providers that build AI systems on general-purpose models and want to report alleged infringements by those model providers.
4. The Digital Omnibus on AI delayed the application of some of the rules that were originally scheduled for today:
The grace period for AI systems already on the market to comply with the watermarking obligations (set out in Article 50.2) was extended to December 2, 2026.
The enforcement of the rules for high-risk AI systems listed in Annex III was postponed to December 2, 2027.
The enforcement of the rules for high-risk AI systems listed in Annex I was postponed to August 2, 2028.
Learn more about the other changes proposed by the Digital Omnibus here.
*To understand the EU AI Act from a practical perspective and prepare for the legal and ethical challenges in AI, join the 31st cohort of my Global AI Policy, Ethics, and Regulation Training Program in September. Register here.