Meta refuses to sign the EU’s AI Code of Practice

Meta declined to sign the European Union code of practice of AI law a few weeks before the blocking rules for providers of the total purpose AI model took effect.

“Europe is on the wrong path of AI,” wrote Joel Kaplan, Meta’s chief global affairs officer, in a LinkedIn post. “We carefully review the European Commission’s Code of Practice for General Purpose AI (GPAI) Models, and Meta will not sign. This code introduces a lot of legal uncertainty to model developers and implements measures well beyond the scope of AI Act.”

The EU Code of Practice – a voluntary framework released earlier this month – aims to help businesses implement processes and systems to comply with the Bloc laws to regulate AI. Among other things, this code requires companies to regularly update documentation on AI tools and services. Prohibits developers who train AI with Pirated Content. Follow the content owner’s request to not use the work in the dataset.

Calling the implementation of EU law “overreach,” Kaplan argued that the law “will narrow down the development and deployment of European frontier AI models and stunts European companies trying to build their businesses on top of it.”

The AI Act, a risk-based regulation for the application of artificial intelligence, completely prohibits “unacceptable” use cases, such as cognitive behavioral manipulation and social scoring. Rules also define a set of “high risk” uses in domains such as education and employment, such as biometric authentication and facial recognition. The law also requires developers to register AI systems and fulfill their risk and quality control obligations.

Tech companies around the world, including people at the forefront of AI races such as Alphabet, Meta, Microsoft and Mistral AI, are urging the Commission to slow its rollout. However, the committee has held it firmly, saying it will not change its timeline.

Also, on Friday, the EU published guidelines for AI model providers ahead of the rules that come into effect on August 2. These rules affect providers of “systemic AI models” such as Openai, Anthropic, Google, and Meta. Companies that have put such models on the market before August 2nd must comply with the law by August 2nd, 2027.