The Code also outlines expectations for GPAI makers to identify and manage systemic risks, provide information about data sources used for AI development, and establish a single point of contact for rights holders to communicate grievances. It also proposes measures for handling serious incidents and substantial systemic risk notifications. The draft Code is expected to undergo significant changes based on feedback and the final version is set to be released on May 1, 2025.
Key takeaways:
- A first draft of a Code of Practice for providers of general-purpose AI models under the European Union’s AI Act has been published, with an invitation for feedback open until November 28.
- The Code of Practice will apply to major AI providers like OpenAI, Google, Meta, and others, and is intended to provide guidance for meeting the EU AI Act’s obligations.
- The Code includes transparency requirements for makers of general-purpose AI models, which are set to enter into force on August 1, 2025, and risk assessment and mitigation requirements for the most powerful models, which are expected to be enforced by August 1, 2027.
- The draft Code also includes measures related to copyright, systemic risk, safety and security, and serious incident reporting, and seeks feedback on these and other areas as it continues to be developed and updated towards a more granular final form for May 1, 2025.