EU's AI Act Code of Practice: Balancing Innovation and Regulation in Latest Draft

EU Releases Third Draft of AI Code of Practice

The European Union has published the third draft of its Code of Practice for General-Purpose AI (GPAI) models, aiming to provide guidance for AI model makers on complying with the EU AI Act 1

1

. This latest iteration, released on March 11, 2025, is expected to be the final revision before the guidelines are finalized in May. The Code addresses key areas such as transparency, copyright, and risk mitigation for powerful AI models.

Streamlined Structure and Refined Commitments

The new draft boasts a more streamlined structure with refined commitments and measures compared to earlier versions. It includes sections on transparency, copyright, and safety and security obligations for the most powerful models with systemic risk (GPAISR) 1

1

. The transparency section provides an example documentation form for GPAIs to ensure downstream deployers have access to key information for compliance.

Industry Concerns and Feedback

Despite efforts to refine the Code, industry stakeholders have expressed "serious concerns" about the latest draft 2

2

. Tech lobby groups, such as CCIA and DOT Europe, argue that some provisions go beyond the requirements of the AI Act and may threaten trade secrets or impose burdensome external risk assessments.

Copyright Issues Remain Contentious

The copyright section remains one of the most controversial areas. The current draft uses terms like "best efforts" and "reasonable measures" when addressing copyright compliance 1

1

. This language has drawn criticism from rights-holders and publishers who argue that it falls short of providing adequate protection. News Media Europe, representing 2,700 news brands, has stated that the draft "raises serious questions about whether no code is better than this code" 2

2

.

Safety and Security Measures

The draft narrows down some previously recommended measures for safety and security, particularly for the most powerful models trained using a total computing power of more than 10^25 FLOPs 1

1

. This adjustment comes in response to feedback from stakeholders.

International Pressure and Regulatory Balance

The EU faces pressure from the U.S. administration, led by President Donald Trump, which has criticized European AI regulations. At the Paris AI Action summit, U.S. Vice President JD Vance warned that overregulation could stifle innovation 1

1

. This international context adds complexity to the EU's efforts to balance innovation with regulation.

Next Steps and Deadlines

Stakeholders have until March 30, 2025, to submit written feedback on the latest draft. The European Commission's AI Office is also working on clarifying guidance that will shape the law's application 1

1

. The final Code of Practice is expected to be released in May, with the AI Act set to be fully applicable by 2027 2

2

.