In the coming days, European users will receive notifications informing them that their public content and interactions with Meta AI may be used to train the group’s AI models.
In a blog post, Meta announced that content shared publicly by European users on Instagram, Facebook, and Messenger can be used to train its artificial intelligence models, as can interactions with its Meta AI chatbot, which has been available in the European Union since March. However, it will be possible to object to the use of your personal data by filling out a dedicated form.
Meta relies on public data to improve its chatbot
On Monday, April 14, 2025, the Californian group announced its intention to train its AI models, such as Llama 4, on comments, captions, or posts “visible to all” on its social platforms, as well as on exchanges with Meta AI. The objective is clear: to enable its conversational agent, which has some weaknesses at this stage, to more easily grasp the “incredibly rich and complex nuances of European communities,” including “dialects, idiomatic expressions, local knowledge, or even the different ways in which each country uses humor or sarcasm” on its platforms.
It’s important to remember that this type of training is neither unique to Meta nor specific to Europe. It’s the method we already use for other regions. We’re following the example of companies like Google and OpenAI, which already use European user data to train their AI, the American group explained.
What Meta Won’t Collect to Improve Its AI Models
In its press release, Meta assures that data exchanged in private conversations—particularly on WhatsApp or Messenger —will not be used for training purposes, nor will data shared publicly by those under 18. Furthermore, it will also be possible to object to the use of one’s personal data.
In the coming days, individuals based in the European Union will begin receiving notifications, in apps or by email, informing them of this new policy. These messages will include a link to a form allowing them to opt out of the use of their data, in accordance with the GDPR.
In May 2024, Meta had already clearly stated its intention to rely on content shared on its platforms for training purposes, which led to an update to its privacy policy. But the American group ultimately backtracked, “pending clarification from regulators,” the press release recalls and then delayed the deployment of its chatbot by several months. “Since then, we have engaged in a constructive dialogue with the Irish DPC and we look forward to continuing to offer the benefits of generative AI to European users,” the group said.
