In a significant development for artificial intelligence (AI) regulation within the European Union, the EU AI Act’s latest draft Code of Practice (CoP) introduces adjusted guidelines aimed at clarifying copyright-related obligations. The refined measures are particularly notable for their proportional framework, aligning compliance requirements with the size and capacity of the individual general-purpose AI model provider. These changes aim to balance innovation with adherence to existing European copyright laws, as indicated in a detailed report by Bloomberg Law.
The draft, published on March 11, 2025, is deemed more feasible than its predecessors, offering legal practitioners a clearer path to establish compliance until harmonized standards are formalized. With penalties for non-compliance including fines reaching up to 3% of the provider’s global turnover, interest and engagement from stakeholders have been robust, with more than 1,000 entities contributing to the development process.
- Providers are urged to document their copyright-relevant commitments thoroughly.
- Measures should be taken to prevent the generation of copyright-infringing outputs from AI systems.
- Providers must respect machine-readable opt-outs from rightsholders for text and data mining activities not covered by the EU’s TDM exception.
These guidelines represent an evolving framework intending to address ongoing uncertainties surrounding AI’s intersection with copyright law. The reliance on recognized protocols like robots.txt for expressing opt-outs is praised for reducing uncertainties, simultaneously fostering a collaborative environment for future standards. Despite progress, questions remain around the prevention of copyright-infringing outputs—a focal point emphasized in the drafting process, highlighting the opportunity for stakeholder feedback until late March 2025.
The article offers a comprehensive overview of the evolving landscape and underscores the CoP as a step forward in regulatory clarity while recognizing the ongoing need for dialogue and refinement. Legal professionals and AI model providers are encouraged to stay abreast of these developments as they prepare for implementation by August 2, 2025.