Google signs EU AI Act code of practice on transparency, joining Microsoft and OpenAI

Google just raised its hand in the EU’s AI regulatory classroom. The tech giant announced on July 30 its intent to join the European Union’s General Purpose AI Code of Practice, a voluntary framework behind the bloc’s sweeping AI Act. What Google actually signed up for The GPAI Code of Practice, finalized around July 10, 2025, is essentially Europe’s rulebook for how the world’s most powerful AI models should behave. It covers documenting training data, ensuring copyright adherence, and mitigating risks for advanced AI systems. Google’s Gemini models fall squarely within its scope. The obligations kick in on August 2, 2025, and they come with specific computational thresholds. Models exceeding 10 to the 23rd FLOPs must meet standard transparency requirements. Models crossing the 10 to the 25th FLOPs threshold get classified as systemic-risk models, which carries even heavier compliance burdens. Google isn’t alone in signing on. The roster includes OpenAI, Microsoft, Amazon, Anthropic, Mistral AI, and IBM. Meta, along with several Chinese AI firms, has declined to join the code. The economic argument Europe is banking on Google has pointed to projections suggesting responsible AI de...