Sexual Deepfakes, the Oversight Board Rebukes Meta: Too Little Protection for Non-Public Figures
The Oversight Board of Meta has ordered the company to remove a video generated with AI from Instagram that sexualized a woman, reversing the decision with which Meta had left it online. The independent committee determined that the content violated the ban on non-consensual intimate images and accompanied the ruling with a series of recommendations to strengthen protections for everyday people—those who are not public figures—against sexual deepfakes.
The case arose from an eight-second video published in September 2025, in which a woman adjusts a tight dress with visible underwear in some frames. The user who brought the content to the committee stated that a friend was impersonated in the video without her consent, and she had already closed her account.
Two users had reported the post to Meta, which had not removed it. One of the two had appealed to the company without getting a response, then turned to the Oversight Board. Only when the committee raised the case did Meta intervene, limiting the content's visibility to adults only, concluding that it did not merit removal according to Community Standards. The company explained that at the time of the report, there were no indications that the person depicted was real because the directly affected individual had not reported the incident: among the elements of absence of consent that Meta acknowledges are the self-reporting of the victim, vindictive contextual captions, requests from law enforcement, or trusted partners.
What Changes for Non-Famous Individuals
In its decision, the committee found that the content met all three criteria set forth in the Adult Sexual Exploitation Policy: non-commercial material in an apparently private context, a woman depicted in conditions of near nudity, and the absence of consent. From this, the removal order and a package of recommendations ensued.
The committee asks Meta to add a new signal of absence of consent: the very fact that the content is a sexualized impersonation generated by AI of a real person, because such impersonation is non-consensual by definition. It also recommends allowing users to designate connected accounts—that is, trusted friends, family, or verified associates—who can report potential abuses on their behalf. Finally, sexual impersonation produced by AI should become a separate category in reporting and appeal forms, distinct from harassment or nudity, and those specialized forms should be available everywhere: currently, only residents of Texas and Florida have access to a form that lists intimate deepfakes among the reporting reasons.
A detail on the internal functioning of moderation also weighs in. Meta’s automated system, which identifies potentially harmful and highly viral content, had intercepted the post the day after its publication but had not forwarded it for priority human review.
The Regulatory Framework and Precedents
The committee places the case in a moving regulatory context: India, the United Kingdom, and Spain are introducing new rules for platforms regarding AI-generated content, the European Union has reached an agreement to ban systems that produce non-consensual sexually explicit material, including “nudification” apps, while X is under examination for explicit images generated by its chatbots. Recommendations already put forward in the past resurface, such as replacing ambiguous policy terms with explicit references to non-consensuality and widely implementing content credentials from the C2PA standard, making provenance data visible to users.
This is not the first time the committee has called out the company: in mid-2025, it deemed the application of Meta's rules inconsistent and unjustifiable, and in March 2026 it asked them to create a rule dedicated to AI content, separate from the one on misinformation, following an investigation into a video showing alleged damage to buildings in Haifa. Meta is required to respond to the recommendations but is not obliged to implement them; if it chooses to adopt them, it will be the committee that monitors their implementation.