Background
Published on 10 July 2025, the General-Purpose AI Code of Practice is a voluntary framework drafted by independent experts to help providers of general-purpose AI (GPAI) models meet key obligations under the EU AI Act on safety, transparency, and copyright.
Chapters
- Transparency: A user-friendly Model Documentation Form to detail model capabilities, limitations, governance, and data provenance (Article 53).
- Copyright: Practical templates and guidance for establishing and documenting copyright-compliance policies covering training data, licensing, and rights management.
- Safety & Security: State-of-the-art risk-management practices, incident-reporting procedures, and robustness measures for systems with systemic-risk obligations (Article 55).
Signature Process
Providers of general-purpose AI models may sign the code by completing a signatory form and send the signed form to the AI Office.
By signing the Code, GPAI providers can demonstrate compliance more efficiently, reduce administrative burden, and gain legal certainty under the AI Act.
Signatories
Leading AI providers such as Amazon, Google, Microsoft, IBM, Anthropic, OpenAI, Cohere, Mistral AI, and Aleph Alpha have already signed the Code.
Details on the Code of Practice can be accessed here.
Source: European Commission
Accompanying this subject we recommend the following content on our website
European Commission publishes template for summary of training data