BRUSSELS, Belgium: EU AI Act enforcement powers for general-purpose models took effect on August 2, allowing the European Commission to investigate providers and impose fines.
The model-provider obligations first began applying on August 2, 2025. They cover general-purpose AI models placed on the European Union market after that date, according to the Commission’s official guidelines.
Providers must maintain technical documents covering model training, testing and evaluations. They must also provide downstream developers with information on each model’s capabilities and limitations.
The rules require companies to adopt policies that comply with EU copyright law. Providers must also publish sufficiently detailed summaries of the content used to train their models.
Developers of models deemed to pose systemic risks face additional requirements. These include adversarial testing, risk assessments, incident reporting and cybersecurity safeguards.
Read: China Proposes Global AI Cooperation Organisation
The Commission can request documents, evaluate models and order corrective measures. Under the EU AI Act, violations can trigger fines of up to €15 million or 3% of annual global turnover, whichever is higher.
The European AI Office will oversee enforcement. It can also consult an independent scientific panel and specialist AI safety firms.
A voluntary code of practice offers companies a route to demonstrate compliance. OpenAI, Google, Microsoft and Anthropic have signed the full code, while Meta has not. Providers of models released before August 2, 2025, have until August 2, 2027, to meet the requirements.
