The European Union has published its first draft Code of Practice detailing how providers of general-purpose AI models must comply with the EU AI Act. According to tech reporter Natasha Lomas at TechCrunch, the 36-page document outlines transparency requirements starting August 2025 and risk assessment measures for powerful AI systems beginning August 2027. The code will affect major AI providers like OpenAI, Google, Meta, and Anthropic.
The draft guidance focuses on several key areas including copyright compliance, systemic risk management, and safety frameworks. Companies must provide information about data sources, establish complaint handling systems, and implement risk assessment procedures for models exceeding specific computing power thresholds. The document is open for feedback until November 28, with the final version expected by May 2025.