EU AI Act: Latest Developments in Guidance for AI Model Makers
As the deadline approaches for finalizing guidance on the compliance of general-purpose AI (GPAI) models with the provisions of the EU AI Act, a third draft of the Code of Practice has been released. This draft, published on March 11, 2025, is anticipated to be the final iteration before the official guidance is adopted.
Overview of the Code of Practice
The Code of Practice is designed to assist GPAI model makers in understanding their legal obligations and in avoiding sanctions for noncompliance. Notably, penalties for breaches of GPAI requirements can reach up to 3% of a company’s global annual revenue.
This latest revision emphasizes a streamlined structure with refined commitments and measures, reflecting feedback from the second draft published in December 2024. The draft is organized into sections that cover commitments and detailed guidance for transparency, copyright, and safety obligations.
Key Areas of Focus
One of the major areas addressed is transparency. The guidance suggests that GPAIs will need to complete a model documentation form, ensuring that downstream deployers of their technology have access to crucial information for compliance.
Another contentious area is copyright. The current draft utilizes terms like “best efforts” and “reasonable measures,” potentially allowing data-mining AI companies to continue acquiring protected information for model training while mitigating risks of copyright infringement.
Safety and Security Obligations
The EU AI Act imposes safety and security requirements specifically on the most powerful models, identified as those with systemic risk. The latest draft narrows some previously recommended measures to streamline compliance.
Pressure from the U.S.
The ongoing discussions surrounding the EU AI Act have not gone unnoticed by the U.S. administration. Criticism of European lawmaking and AI regulations has emerged, with U.S. officials warning that overregulation could hamper innovation. This backdrop adds pressure for the EU to ease requirements amidst lobbying efforts from American tech firms.
Future Implications
As the final guidance is prepared, the European Commission is simultaneously producing additional clarifying documents to define GPAIs and their responsibilities. Stakeholders are advised to stay tuned for further updates that may shape the operational landscape for AI developers in Europe.
The outcomes of these discussions and the implementation of the Code will likely have profound implications for the future of AI governance, balancing innovation with regulatory compliance.
