EU Releases First Draft Guidance to Help Major AI Models Comply With New Rules
The European Union publishes a preliminary Code of Practice to guide general-purpose AI providers like OpenAI and Google in meeting the requirements of the AI Act. The draft seeks public feedback to shape final compliance measures ahead of upcoming regulatory deadlines.
The European Union releases a first draft of a Code of Practice designed to help providers of general-purpose AI models comply with the bloc's AI Act. The 36-page draft serves as a high-level outline of guiding principles and targets major AI companies like OpenAI, Google, Meta, and Anthropic. These companies use the code as a roadmap to avoid enforcement actions under the risk-based regulatory framework that takes effect over the coming years.
The drafting committee actively invites feedback from industry and civil society until November 28 to help resolve a series of open questions. Because the current document is light on specific details, this public input plays a crucial role in shaping the final substance, sub-measures, and key performance indicators. While the code offers best practice suggestions, GPAI providers retain the option to deviate if they demonstrate compliance through alternative measures.
Compliance deadlines create a staggered timeline for AI companies operating in Europe. Transparency requirements for all general-purpose AI models take effect on August 1, 2025, while the most powerful models classified as posing systemic risk face stricter risk assessment rules starting August 1, 2027. The draft assumes only a small number of AI makers and models fall into this highest-risk category.