Meta Oversight Board slams inadequate deepfake safeguards

The Meta Oversight Board has criticized Meta’s current content moderation policies as ‘consistently and fundamentally inadequate’ to address the rapid rise of AI-generated deepfakes. The board issued binding rulings to remove two specific manipulated videos that had previously been left online by Meta.

The first case involved a deepfake video of a Scottish Labour Party councillor. The video used synthetic audio and facial manipulation to falsely depict the politician making inflammatory, hateful comments regarding refugees. While Meta initially declined to remove the content, arguing it did not violate specific rules or involve election interference, the Oversight Board ruled it violated hate speech policies.

The second case involved a manipulated video of a Muslim woman activist. The video was digitally altered to mock her work as a health advocate by portraying her in a way that undermined her professional credibility.

In response to these gaps, the Oversight Board has recommended nine policy changes. These include creating a dedicated community standard for generative AI, implementing ‘High Risk’ labels for potentially synthetic content, and treating AI-generated impersonations of private individuals as non-consensual by default. The board noted that deepfakes are increasingly used to harass and silence women in public life.