Meta is changing how it handles some Facebook and Instagram user data in Europe for AI training. The company is notifying users through in-app messages about privacy policy changes that take effect June 26.
The key point for users from the EU is simple: there is an opt-out path, but it is not presented as a one-click switch. Meta asks users to justify the request by explaining how the company’s use of their data affects them personally.
What Meta says it wants to use
Meta says it has a "legitimate interest" in using certain user data to train its AI models. The data named in the source includes posts, comments, audio, and messages to businesses.
The company draws a line around some communications. Private messages to friends and family are excluded from this AI training use, according to the source article.
That exclusion does not cover every way a person may appear on Meta’s platforms. Mentions of your name are not excluded. Pictures of other users that show you are also not excluded.
For users, that distinction matters because data can involve more than what someone deliberately posts on their own account. A person may appear in another user’s image or be named in someone else’s content, and the source indicates those cases are not carved out in the same way as private messages to friends and family.
How the opt-out works
EU users can opt out through Facebook or Instagram, but Meta requires more than a basic preference. The request must explain how Meta’s use of the user’s data affects that person personally.
The source notes that a reference to the GDPR should suffice here. Meta says it will stop using a user’s data after a positive review of the request.
That review step is central. The opt-out is available, but Meta’s process still appears to place the user’s objection inside a request system rather than treating it as an automatic refusal.
- Who is affected: Facebook and Instagram users from the EU.
- What is changing: Meta’s privacy policy in Europe, effective June 26.
- What data is named: posts, comments, audio, and messages to businesses.
- What is excluded: private messages to friends and family.
- What users must do: explain how the data use affects them personally.
Why this is part of a larger AI data issue
Meta’s use of platform data for AI training is not described as new. The company has also trained the Meta AI assistant using publicly accessible texts and images from Facebook and Instagram.
According to the source, texts were fed into the Llama language model, while images were used for the Emu AI image generator. Meta’s chief lobbyist Nick Clegg also said that data entered into Meta AI can be used for AI training.
Taken together, these details show that Meta’s AI strategy is tied closely to the content and interactions available across its services. The latest policy change makes that relationship more visible to users in Europe because the company is sending in-app notifications and offering an objection process.
The issue is not only whether AI systems can be trained on platform data. It is also about what control users have when their posts, comments, images, business messages, or appearances in other people’s content may become part of that training process.
Regulators are likely to examine the approach
The source article says EU authorities will probably look into whether Meta has a valid "legitimate interest" and whether the opt-out procedure is set up properly.
That expected scrutiny fits a broader concern around AI and personal data. A new interim report by the EU data protection authorities still views OpenAI’s ChatGPT critically, especially where personal data is entered in prompts and where personal data must be deleted from models.
The source also states that operators are responsible for both issues. Technical difficulties are not an excuse for breaking data protection rules.
For Meta, the practical question is whether its review-based objection process will be considered sufficient. For users, the immediate question is whether they want to file an opt-out request and how clearly they can explain the personal effect of Meta using their data for AI training.
What users should understand now
The most important takeaway is that Meta is not saying all European user data is outside AI training by default. It is notifying users of a policy change, naming categories of data it wants to use, and offering an opt-out process that requires a personal explanation.
Private messages to friends and family are excluded, but that does not settle every privacy concern. Mentions, public-facing activity, messages to businesses, and images posted by others can still raise questions for users who do not want their data connected to AI model training.
The June 26 effective date gives the change a clear timeline. The regulatory response may determine whether Meta’s reliance on "legitimate interest" and its opt-out design hold up under closer examination.