Meta has said it will not join the voluntary Code of Practice for Artificial Intelligence, which was developed by the European Commission ahead of the AI Act, which the company believes creates legal risks and limits innovation.
Meta has said it will not join the voluntary Code of Practice for Artificial Intelligence, which was developed by the European Commission ahead of the AI Act, which the company believes creates legal risks and limits innovation.
According to TechCrunch, in a LinkedIn post, Meta’s vice president of global affairs, Joel Kaplan, explained the company’s decision by saying that the European approach to AI regulation is overly strict and goes against the interests of developers. According to him, the Code contains requirements that go beyond the law itself and also creates «legal uncertainty.»
The document, presented by the European Commission in early July, was supposed to be the first stage of companies' adaptation to the AI Act — a comprehensive law that for the first time in the world systematically regulates artificial intelligence. Although the code is voluntary, it was created as a basis for compliance with future legal obligations. In particular, it prohibits training models on pirated content, requires technical documentation for models and gives authors the right to prohibit the use of their works in training samples.
Meta believes that such restrictions could hold back the development of advanced AI models in Europe. Kaplan stressed that it would «stifle innovation» and become an obstacle for local startups that want to work with new technologies. According to the company, the code could put the EU at a disadvantage in the global race for leadership in AI.
The AI Act itself will come into force on 2 August 2025. It prohibits some of the most dangerous practices, such as social scoring or behavioural manipulation, and identifies «high-risk» areas, including biometrics, education and employment. Developers of general-purpose models (such as GPT, Gemini or LLaMA) will be required to undergo certification, provide technical descriptions and guarantee compliance with security requirements. For models released before August 2025, the adaptation period will last until 2027.
Meta is not the only company criticizing the EU’s approach. Back in June, representatives of OpenAI, Google, Anthropic and others urged the European Commission to delay the introduction of binding rules. However, Brussels remained adamant: European officials believe that transparent and clear rules will make the development of AI safe and ethical in the long term.
We previously wrote about how Mark Zuckerberg announced that Meta would spend hundreds of billions of dollars building several huge AI data centers.
