Meta Oversight Board demands stronger safeguards against sexualised deepfakes for private users

Meta’s Oversight Board has issued a series of recommendations urging the technology giant to strengthen protections for non-public figures against sexualised AI deepfakes. The independent body argues that Meta’s existing content moderation frameworks disproportionately favour public figures, who possess greater access to the resources required to prove non-consent, such as law enforcement involvement or media coverage.

The recommendations follow a specific investigation into an incident where Meta failed to remove a sexualised deepfake of a private individual from Instagram, despite multiple reports from friends and the victim. The Board overturned Meta’s decision to keep the post online and mandated its removal, noting that the company had initially retained the content because it lacked indication the individual was a real person.

In its report, the Board highlighted that Meta’s reliance on indicators such as self-reporting, law enforcement reports, or media involvement makes it exceptionally difficult for ordinary users to establish non-consent. The Board asserted that AI-generated impersonations should be considered non-consensual by default and recommended adding them explicitly to Meta’s Adult Sexual Exploitation policy.

To address these gaps, the Board suggested allowing users to designate "connected accounts," such as trusted friends and family, to report potential violations on behalf of victims. It also urged Meta to create a separate reporting category for AI-generated sexual impersonation, distinct from harassment and nudity, and to make this option available to all users globally rather than limiting it to residents of Texas and Florida.

The Board noted that the proliferation of AI tools has led to a global increase in sexualised non-consensual content, causing significant reputational and psychological harm, particularly to women and girls. This marks the second time in recent months the Board has criticised Meta’s approach to AI content, having previously called for a distinct AI content rule separate from misinformation policies in March 2025.

Meta is required to respond to these recommendations but is not obligated to implement them. If the company chooses to adopt the changes, the Board will monitor the implementation. The recommendations come amid broader concerns about the scale, speed, and sophistication of AI tools, which the Board warns have a chilling effect on participation in social and political life.