EU AI Office gains enforcement powers for GPAI models

The European Commission is gaining formal enforcement powers regarding General Purpose AI (GPAI) models under the AI Act. While obligations for new GPAI models began on August 2, 2025, the Commission’s ability to fully enforce these rules—including requesting documentation, conducting evaluations, and imposing sanctions—becomes effective on August 2, 2026.

Enforcement tools include technical dialogues for compliance, but the Commission also holds the authority to demand model access (via API or source code) and implement risk-mitigation measures. In severe cases, penalties can reach up to 15 million euros or 3% of a company’s total global annual turnover. The Commission may also require the restriction or withdrawal of specific models.

In parallel, a Code of Practice regarding the labeling of AI-generated content has been established. Approximately 190 organizations, including Anthropic, Google, Meta, Microsoft, OpenAI, and Mistral, have joined the code. This framework requires providers to implement rigorous testing, verification, and monitoring of technologies such as watermarking, metadata, and digital fingerprinting to ensure the effectiveness of AI content detection.