Meta’s Oversight Board is pushing the company to overhaul how it handles sexualized deepfakes, saying its current system leaves private individuals with almost no real way to protect themselves. The Board has released a set of recommendations after investigating a case where Meta repeatedly refused to remove an AI-generated intimate video of a private woman posted on Instagram, even after multiple reports from users and a formal appeal.
The case that triggered the investigation involved a user who reported a video showing a woman adjusting her dress, with her underwear visible in several frames. The person in the video had already deleted her own Instagram account. Two users reported the video. Meta did not remove it. One of those users then appealed to the Oversight Board. Meta still did not remove it. Only after the Board itself raised the issue directly with Meta did the company act, and even then, it only made the post adults-only rather than taking it down. The Board has since overruled that decision and required Meta to remove the video.
According to Engadget, the Board’s investigation exposed a structural problem in how Meta determines whether intimate content is non-consensual. Under Meta’s current rules, the most reliable way to establish non-consent is for the person depicted to report the content themselves. Other accepted indicators include reports from law enforcement, media outlets, or trusted partners, or captions that suggest the content is being shared in a vengeful or sensationalist way.
The problem is obvious. Those options are far more accessible to public figures than to private individuals. A celebrity or politician can realistically get media coverage or law enforcement involved. Most people cannot. The Board says this creates a two-tier system where ordinary people are effectively left to fend for themselves.
To fix this, the Board has put forward several concrete recommendations for Meta:
- Add AI-generated sexual impersonation to Meta’s Adult Sexual Exploitation policy, treating such content as non-consensual by default
- Allow users to designate trusted contacts, such as friends or family, who can report violations on their behalf
- Create a separate reporting category specifically for AI-generated sexual impersonation, distinct from harassment or nudity
- Make the specialized deepfake reporting form available to all users globally, not just people in Texas and Florida
That last point is particularly striking. Right now, only residents of two US states have access to a reporting form that lists deepfake intimate imagery as a specific reason for reporting. Everyone else has to work around categories that were not designed with this type of content in mind. The Board’s position is straightforward: this is a global problem and the tools to address it should be available globally.
Meta is required to respond to these recommendations but is not obligated to act on them. If it does choose to implement any of them, the Board says it will monitor the rollout.
This case fits into a broader pattern of the Oversight Board pushing back on how Meta handles AI-generated content. In mid-2025, the Board criticized Meta’s content moderation as “incoherent and unjustifiable” when it came to AI material. Before that, in March 2025, it urged Meta to create a dedicated policy for AI content separate from its existing misinformation rules, following an investigation into a fake video that falsely showed damaged buildings in an Israeli city. That video had been posted by an account posing as a news outlet but was actually run by a user based in the Philippines.
The broader context matters here. AI tools have made it faster, cheaper, and easier to create convincing fake intimate imagery of real people. The Board noted in its report that the spread of sexualized deepfakes causes serious reputational and psychological harm, affects women and girls disproportionately, and discourages people from participating in public life. Meta’s current policies were not built with that reality in mind, and the gap between what the rules cover and what people actually face is getting wider.