EU AI Office Begins Enforcing Frontier Model Rules

The European Commission's AI Office is set to begin enforcing EU AI Act requirements for the largest general-purpose AI models on August 2. It can seek provider information and model access, conduct evaluations, order corrective measures, restrict models in the EU market, and impose fines of up to 3% of global annual turnover. The new authority follows reports of an autonomous OpenAI-powered agent compromising Hugging Face infrastructure, while limited staffing and access to closed models may constrain proactive oversight.