EU rules governing general-purpose AI (GPAI) models have officially taken effect, introducing stricter requirements for transparency, safety, and accountability in AI systems.
From August 2, providers must disclose how their models are trained, respect copyright protections, and follow responsible development practices. The Commission has issued guidance and a template to help companies comply, while also confirming that adherence to the voluntary GPAI Code of Practice can ease compliance obligations.
More advanced models with systemic risks — those trained with over 10^25 FLOP — will face additional requirements, including safety and security measures and mandatory notification to the Commission. Providers of models already on the market before August 2025 will have until August 2027 to meet the new standards.
