EU Rolls Out AI Code With Broad Copyright, Transparency Rules

In a significant move, the European Union has introduced a code of practice aimed at assisting companies in adhering to its pioneering AI Act. This new code emphasizes copyright protections for creators and mandates transparency for advanced AI models. According to the European Commission, developers will be required to maintain updated documentation detailing their AI systems' functionalities. This documentation will be accessible to regulators and third parties interested in integrating AI into their products. Additionally, companies are prohibited from utilizing pirated materials for training AI and must comply with requests from writers and artists to exclude copyrighted works from datasets. If an AI system generates content that violates copyright laws, firms will need to implement a procedure to rectify such infringements. While the code of practice is voluntary, it aims to assist companies in setting up internal strategies for compliance with the AI law. The regulation is set to be implemented on a staggered basis, with specific rules governing general-purpose AI, such as OpenAI's ChatGPT and Anthropic's Claude, coming into effect next month. Companies that violate the AI Act could face fines of up to 7% of their annual revenue or 3% for those developing advanced AI models. This code has sparked controversy and discontent among some technology firms, including Meta Platforms Inc. and Alphabet Inc., who argue that earlier drafts imposed excessive regulations that exceeded the original scope of the AI Act. Recently, prominent European companies like ASML Holding NV, Airbus SE, and Mistral AI have urged the commission to delay the AI Act's implementation for two years, advocating for a more 'innovation-friendly' regulatory framework. Despite missing an initial deadline in May for publishing the code of practice, the commission has so far resisted calls for postponement. The code was created with input from various stakeholders, including AI labs, tech companies, academic institutions, and digital rights advocates. Full enforcement of the AI Act will not begin until August 2026, with national courts handling enforcement in the interim, despite potentially lacking specialized technical knowledge. Signing onto the code of practice is expected to provide companies with greater legal certainty moving forward.