Meta's Oversight Board Ordered Removal of an AI-Generated Sexualized Deepfake on Instagram and Called for Major Policy Reforms

Key Policy Changes the Board Called On Meta to Adopt

The Board's ruling went beyond the individual case and issued several binding and non-binding recommendations to overhaul how Meta handles AI-generated sexualized content, particularly for non-public figures . The key changes include:

Treat AI-Generated Impersonation as Non-Consensual by Default

The Board explicitly stated that AI-generated impersonations in sexualized contexts should be considered "non-consensual by default" and recommended adding these to Meta's Adult Sexual Exploitation policy . This shifts the burden away from victims having to prove non-consent and onto the platform to recognize the inherent harm of such content .

Allow Trusted Contacts to Report on Behalf of Victims

A significant procedural change recommended by the Board is allowing users to designate "connected accounts" — such as trusted friends or family members — who can report potential violations like non-consensual intimate imagery on behalf of the victim . This addresses a key gap where victims may be unaware of the content or unable to report it themselves due to trauma or fear.

Create a Global Reporting Category for AI-Generated Sexualized Imagery

The Board called on Meta to create a dedicated, global reporting category for AI-generated non-consensual sexualized imagery, moving beyond the company's current classification under a generalized bullying and harassment rule . This would make reports easier to file and track.

Update Terminology from "Derogatory Sexualized Photoshop" to "Non-Consensual" Manipulated Media

The Board has previously criticized Meta's outdated terminology, which used language like "derogatory sexualized photoshop" . The Board recommended updating Meta's rules to use clearer, more accurate terms that reflect the modern reality of AI-generated abuse, emphasizing the non-consensual nature of the content rather than the method of creation .

Implement Content Credentials at Scale and Improve Detection

In a related AI-content case, the Board also called on Meta to implement scalable tools for identifying AI-generated content, including adopting the C2PA Content Credentials standard — a provenance standard for tracking the origin and modifications of media . The Board found Meta's current systems rely too heavily on self-disclosure by users and lack robust automated detection, particularly for video and audio .

Focus on Protecting Non-Public Figures

A central theme of the ruling was the protection gap for ordinary, non-public people. The Board emphasized that non-public figures targeted by sexualized deepfakes are especially harmed by systems that require them to prove non-consent themselves, because they often lack the public visibility or institutional support available to public figures . The case itself centered on an AI-generated sexualized impersonation of a woman, reported by a friend who stated the content was non-consensual and reputationally damaging .

Addressing Gender-Based Harm

The Board's decision also acknowledged the disproportionate harm these deepfakes cause to women and girls. By calling for stronger automated detection, a shift in the burden of proof, and clearer policies, the Board aimed to address the systemic nature of gender-based online abuse .

Next Steps for Meta

Meta is required to respond to the Oversight Board's recommendations within 60 days . While the Board's recommendation to remove the specific video is binding, its broader policy recommendations are not binding but carry significant public and reputational weight . The ruling builds on previous Board decisions in 2024 and 2025 that found Meta's deepfake policies to be "incoherent" and "not sufficiently clear" .