Meta has chosen not to sign the European Union’s code of practice for general-purpose AI models, setting up another public clash between a major technology company and the bloc’s AI regulatory agenda.
The decision comes weeks before new rules for providers of general-purpose AI models are due to take effect. The code is voluntary, but it is designed to guide companies as they prepare systems and processes for complying with the AI Act.
Why Meta Is Refusing The Code
Meta’s position was laid out by Joel Kaplan, the company’s chief global affairs officer, in a post on LinkedIn. He said Meta had reviewed the European Commission’s Code of Practice for general-purpose AI (GPAI) models and would not sign it.
"Europe is heading down the wrong path on AI,"
Kaplan argued that the code creates uncertainty for companies building AI models. He also said it includes measures that, in Meta’s view, extend beyond the AI Act itself.
"We have carefully reviewed the European Commission’s Code of Practice for general-purpose AI (GPAI) models and Meta won’t be signing it. This Code introduces a number of legal uncertainties for model developers, as well as measures which go far beyond the scope of the AI Act."
Meta’s objection is not only procedural. Kaplan described the EU’s implementation of the law as "overreach" and said it would "throttle the development and deployment of frontier AI models in Europe and will stunt European companies looking to build businesses on top of them."
That framing puts Meta’s refusal in the broader debate over whether AI regulation can create predictable safeguards without slowing companies that develop or rely on advanced models. The company is presenting the code as a burden on model developers and on European businesses that may want to build products using those models.
What The EU Code Asks Companies To Do
The EU’s code of practice was published earlier this month as a voluntary framework. Its purpose is to help companies prepare for the AI Act’s requirements, especially where general-purpose AI models are involved.
According to the source article, the code includes several practical obligations for companies. It asks them to provide and regularly update documentation about AI tools and services. It also bans developers from training AI on pirated content.
The code also addresses the role of content owners. Companies must comply with requests from content owners who do not want their works used in datasets.
Those requirements explain why the code matters even though it is voluntary. For companies developing large AI models, documentation, training data practices, and requests from rights holders can shape how products are built, maintained, and offered in the market.
For regulators, the code appears intended to turn broad legal expectations into more concrete operating practices. For model developers, Meta’s response shows that the same guidance can be seen as adding legal and operational questions rather than resolving them.
How The AI Act Frames Risk
The AI Act is described as a risk-based regulation for applications of artificial intelligence. That means it does not treat every AI use case the same way. Instead, it draws lines around uses considered more dangerous or sensitive.
Some "unacceptable risk" uses are banned outright. The source article identifies cognitive behavioral manipulation and social scoring as examples.
The law also defines "high-risk" uses. These include biometrics and facial recognition, as well as uses in areas such as education and employment.
For developers, the AI Act also brings administrative and management duties. It requires developers to register AI systems and meet risk- and quality-management obligations.
This is the legal setting for the dispute over the code of practice. The AI Act sets the regulatory structure, while the code is meant to help companies understand and implement compliance processes. Meta’s criticism is that the code does more than that.
Tech Companies Are Still Pushing Back
Meta is not the only company resisting the EU’s AI rules. Tech companies from across the world, including Alphabet, Meta, Microsoft, and Mistral AI, have been fighting the rules.
Some companies have urged the European Commission to delay the rollout. The Commission, however, has held firm and said it will not change its timeline.
That timing is important. The EU also published guidelines for providers of AI models ahead of rules that will go into effect on August 2.
Those rules would affect providers of "general-purpose AI models with systemic risk," including OpenAI, Anthropic, Google, and Meta. Companies with such models already on the market before August 2 will have to comply with the legislation by August 2, 2027.
The result is a regulatory countdown with unresolved tension. The EU is moving ahead with its schedule, while some of the most prominent AI companies argue that the framework could create uncertainty or slow deployment in Europe.
Meta’s refusal to sign the code does not stop the AI Act timeline. It does, however, make clear that voluntary compliance tools can become part of the conflict when companies believe they expand the practical reach of a law.
What To Watch Next
The next phase will center on how companies handle the August 2 rules and how regulators apply the guidance around general-purpose AI models. The source article does not say that Meta is refusing to comply with the AI Act itself; it says Meta will not sign the voluntary code of practice.
That distinction matters. A company can reject voluntary guidance while still facing binding legal obligations under the AI Act. The practical question is whether refusing the code makes compliance more difficult, more contested, or simply more company-specific.
For the AI industry, the dispute shows how implementation details can become as important as the law’s headline goals. Documentation, training data limits, rights-holder requests, registration, and risk management are not abstract issues for model providers. They affect how AI systems are developed, maintained, and launched.
For Europe, the Commission’s decision to hold its timeline keeps pressure on companies to prepare. For Meta, the refusal signals that the company wants to challenge the shape of the implementation, not just the concept of AI regulation.