Meta trains AI on users' posts and images – how to opt out

Meta has been using ordinary users' posts and images on Facebook and Instagram to train its generative AI services since May 2025. Users can opt out, but Norway's Data Protection Authority (Datatilsynet) says it is almost impossible to be sure your personal data is not used anyway.

Tobias Judin, head of the authority's international section, says Meta cannot do whatever it likes with this content, because personal data is regulated by law. As a rule of thumb, users must give voluntary consent, meaning they say yes first, before a company uses their data for AI training. For less sensitive data, or for AI models serving weighty purposes, offering an opt-out may be enough. Meta argues it does not need consent and that an opt-out is sufficient. Judin calls this controversial, and Meta has been reported to the Irish data protection authority over the practice. That case is still ongoing.

To opt out, a user fills in an objection form after logging in with a Facebook or Instagram account. Datatilsynet has a page that serves as a guide. Giving a reason is optional, and Meta has said it accepts all objections with or without one. The process takes about a minute.

The form only covers content you post yourself. Content that other people post about you will still be used for AI training unless they have also submitted the form. Judin says this is why you cannot be certain your personal data stays out of training, even after filling in the form or deleting your accounts.

Separately, several Norwegian media outlets recently won in district court against Meta over planned changes to the terms for business accounts. The outlets argued the terms would let Meta scrape every article, image and video they publish, including paywalled content. Meta is barred from implementing the change until its legality is finally settled.