Meta refuses to sign EU’s AI code of practice

In a significant move, Meta has declined to endorse the European Union's recently proposed code of practice associated with its AI Act, just weeks before the new regulations are set to come into force. Joel Kaplan, Meta’s chief global affairs officer, expressed his concerns in a LinkedIn post, asserting that "Europe is heading down the wrong path on AI." Kaplan elaborated that after a thorough evaluation of the EU's Code of Practice for general-purpose AI (GPAI) models, Meta has decided against signing it. He highlighted that the code introduces numerous legal ambiguities for developers and imposes requirements that exceed the intended scope of the AI Act. The voluntary framework, published earlier this month, is designed to assist companies in aligning their practices with the EU's legislative measures for AI regulation. Key provisions include mandates for companies to maintain and update documentation regarding their AI tools, a prohibition against using pirated content for training AI, and an obligation to respect content owners' requests to exclude their works from datasets. Kaplan criticized the EU's approach as "overreach," warning that such legislation could hinder the advancement and implementation of cutting-edge AI technologies in Europe. He believes that it could also impede European firms aiming to innovate in the AI sector. The AI Act encompasses risk-based regulations, outright banning certain high-risk applications, including manipulative behavioral techniques and social scoring. It categorizes various uses of AI as "high-risk," particularly in areas like biometrics and facial recognition, as well as in education and employment contexts. Developers are also required to register their AI systems and adhere to risk and quality management protocols. Tech giants, including Meta, Alphabet, and Microsoft, have been vocal about their opposition to these rules, even calling for a postponement of the rollout. However, the European Commission remains steadfast in its timeline, refusing to alter the implementation schedule. Additionally, the EU has released guidelines for AI model providers ahead of the upcoming regulations, which will impact companies offering general-purpose AI models with systemic risks, such as OpenAI and Google. Businesses with such models on the market will need to comply by August 2, 2027.