The Meta Oversight Board has released urgent recommendations to improve protection for ordinary users against sexualized deepfakes generated by artificial intelligence. Currently, Meta's system favors public figures, leaving regular people without effective tools to remove non-consensual content. The Board calls for concrete changes to the Adult Sexual Exploitation policy, including explicitly adding AI-generated impersonations as automatic violations without requiring proof of non-consent.
Complicated reporting for ordinary people
The issue emerged from a specific case: a user repeatedly reported a video on Instagram showing a woman adjusting her dress, with her underwear visible in some frames. The video was AI-generated and the real person had already closed her account. Two initial reports were ignored by Meta, and a subsequent appeal did not lead to removal. Only after the Board's intervention did Meta restrict the post to adults only, but it initially did not remove it. This case demonstrates how difficult it is for those without media visibility to obtain justice.
Sponsored Protocol
Key recommendations from the Board
The Board proposes three main measures. First: add AI-generated impersonations to the Adult Sexual Exploitation policy, considering them non-consensual by default. Second: allow users to designate trusted accounts such as friends or family who can report violations on their behalf. Third: create a separate category for AI sexual impersonations in reporting and appeal forms, distinct from harassment and nudity. Currently only users in Texas and Florida have access to a specialized form for intimate deepfakes; the Board wants it extended globally because the problem is worldwide.
Sponsored Protocol
A system that penalizes non-famous individuals
Meta justified the non-removal by stating there were no indications that the person in the video was real. Under current rules, only self-reporting by the victim, law enforcement, media, or trusted partners count as evidence of non-consent. But for an ordinary person, it is difficult to involve police or journalists. The Board emphasizes that this approach favors celebrities and leaves ordinary citizens defenseless, with a disproportionate impact on women and girls. The proliferation of sexual deepfakes causes severe reputational and psychological harm, and the Board describes Meta's current inability to enforce rules consistently as incoherent and unjustifiable.
This is not the first time the Board has criticized Meta on AI content moderation. In March 2025, it already recommended separating policies for AI content from those on misinformation, following a case of a fake video about Haifa. Now, as technology advances, pressure increases. Meta is required to respond to the recommendations but is not obligated to implement them. However, the Board will monitor any actions. For the specific case, the Board overturned Meta's decision and required removal of the video.
Sponsored Protocol
For more on AI security, read about OpenAI's Patch the Planet initiative for open source security. Also explore how AI infrastructure is evolving with Microsoft and Chevron's power plant. For general background on deepfakes, see Wikipedia.
