Meta’s plan to use EU user data for AI training is facing a renewed challenge from privacy watchdog Noyb, which says the company is putting too much burden on users who do not want their public posts included in training data sets.
The dispute centers on whether Meta can rely on an opt-out system, whether earlier objections still count, and whether the company can technically keep objecting users’ data out of its AI systems once training begins.
The opt-out dispute
Noyb sent Meta a cease-and-desist letter on Wednesday, threatening to pursue a potentially billion-dollar class action to block Meta’s AI training in the European Union. According to the source article, the training starts soon in the EU, and Meta recently notified users that they had until May 27 to opt their public posts out of Meta’s AI training data sets.
Noyb’s core concern is that some users who objected in 2024 may have to object again. The group says Meta is requiring users who already opted out of AI training in 2024 to opt out again or risk losing their chance to keep their data out of Meta’s models. Noyb argues that training data likely cannot be easily deleted once it is used.
That claim goes directly to the trust question around AI training. If a user already exercised an objection right, Noyb says the process should not reset in a way that forces the same person to act again under time pressure.
Meta disputes that interpretation. In a blog, Meta said it would honor all objection forms it had already received, along with newly submitted ones.
Why Noyb says consent matters
Noyb argues that Meta could avoid this conflict by asking for consent rather than requiring users to opt out. Meta, however, plans to use a “legitimate interest” legal basis for collecting AI training data. Meta has said that basis follows the clear guidelines of the European Data Protection Committee of December 2024 and reflects consensus between EU data protection authorities.
Noyb Chairman Max Schrems rejects that reasoning. He said the European Court of Justice has already held that Meta cannot claim a “legitimate interest” in targeting users with advertising, and he questioned why the company should be able to claim that basis for AI training.
The disagreement is not only legal. It is also about default settings. Under Meta’s approach, users who do not want their data used must take action. Under the consent-based model Noyb favors, Meta would need users to actively agree first.
That difference matters because AI training is difficult to reverse. Noyb’s warning is that once personal data has been processed into training systems, a later objection may not offer the same practical protection.
The technical concern behind the legal fight
Noyb also questions whether Meta can reliably separate users who objected from users who did not. The watchdog pointed to Meta’s previous argument, made in relation to EU-US data transfers, that a social network is a single system where EU and non-EU users cannot always be cleanly differentiated because many nodes are shared.
For Noyb, that raises a practical problem. If Meta’s systems contain shared interactions, the company may struggle to keep objecting users’ data out of AI training while still using data from users who did not object.
Noyb warned that messages involving one user who objected and another who did not could end up in Meta’s AI systems despite the first user’s objection. The group also criticized Meta’s plan as including data that users may not think of as public, such as disappearing stories usually seen by small audiences.
That distinction is central to Noyb’s argument. The group says this is different from AI crawlers collecting material posted on a public website, because some social platform content may be technically visible in a limited way without feeling broadly public to the user.
Meta’s defense
Meta says it has given EU users a clear way to object to their data being used for training AI at Meta. A company spokesperson said users were notified by email and in-app notifications and that they can object at any time.
The spokesperson also criticized “Noyb’s copycat actions” as part of an attempt by activist groups to delay AI innovation in the EU. Meta argues that such delays harm consumers and businesses that could benefit from advanced technologies.
In a blog post, Meta said training on EU users is important for building AI tools for Europeans. The company pointed to the need for models informed by “dialects and colloquialisms,” “hyper-local knowledge,” and the ways different countries use humor and sarcasm on its products.
Meta also says its EU AI training effort is more transparent than the work of competitors Google and OpenAI. According to Meta, those companies have already used data from European users to train AI models, while Meta says it has taken steps to inform users.
What happens next
Noyb requested a response from Meta by May 21. The group says it is evaluating options to seek injunctive relief in various jurisdictions and could pursue a class action worth possibly “billions in damages.” Noyb says the goal is to protect the data rights of 400 million monthly active EU users.
Meta, meanwhile, says its plan follows extensive and ongoing engagement with the Irish Data Protection Commission. It also says the approach reflects consensus among EU Data Protection Authorities.
Noyb disputes that picture. The group argues that national DPAs have “largely stayed silent on the legality of AI training without consent” and says Meta appears to have moved ahead anyway.
The result is a clear clash over how AI companies should treat social platform data. Meta frames the issue as necessary AI development for Europe. Noyb frames it as a consent problem, arguing that users should not have to race through another opt-out process to keep personal data away from AI training.