EU rules on general-purpose AI models start to apply tomorrow, bringing more transparency, safety and accountability

This will bring more transparency, safety and accountability to AI systems on the market. These rules mean clearer information about how AI models are trained, better enforcement of copyright protections and more responsible AI development.

To assist providers, the Commission published guidelines clarifying who must comply with the obligations. GPAI models are defined as those trained with over 10^23 FLOP and capable of generating language. Additionally, the Commission published a template to help providers in summarising the data used to train their models.

The Commission and the Member States have also confirmed that the GPAI Code of Practice, developed by independent experts, is an adequate voluntary tool for providers of GPAI models to demonstrate compliance with the AI Act. Providers who sign and adhere to the Code will benefit from reduced burden and increased legal certainty.

From tomorrow, providers must comply with transparency and copyright obligations when placing GPAI models on the EU market. Models already on the market before 2 August 2025 must ensure compliance by 2 August 2027.

Providers of the most advanced or impactful models presenting systemic risks — i.e. those exceeding 10^25 FLOP — will need to meet additional obligations, such as notifying the Commission and ensuring the model's safety and security.

More information is available in the enforcement timeline for GPAI model. A factpage with further details is also available online.

Find further information:

• AI Act
• Commission Opinion on the assessment of the General-Purpose AI Code of Practice
• GPAI Code of Practice
• Drawing-up a General-Purpose AI Code of Practice
• AI Board