Meta decided to remove AI-generated videos targeting two women, including a British politician, after the Oversight Board criticized the company's protection measures against deepfakes, calling them 'fundamentally inadequate.'
Earlier on Thursday, the Oversight Board, a quasi-independent body funded by the company to review moderation decisions, demanded that Meta remove fabricated videos from its platforms. The Board noted that such cases undermine Meta's own policies regarding hate speech, disinformation, and harassment.
The first AI-generated video depicted a Labour Party member in Scotland making provocative statements about refugees. No public record existed to confirm the politician's remarks, and the video showed clear signs of forgery, including audio not synchronized with her facial movements.
The second fabricated video was aimed at a Muslim activist in Europe. The goal was to discredit her role as a health advocate by falsely showing her consuming fast food while giving advice on mindful eating and performing absurd exercises.
The Board stated that Meta's policies are 'consistently and fundamentally inadequate' to counter the growing flow of AI-generated content on its platforms, which include Facebook and Instagram.
Co-chair Pamela San Martin stated in her submission: 'From politicians to private citizens, AI-generated deepfakes are increasingly used to harass and silence women who participate in public discourse.' She added that 'these cases demonstrate a broader, troubling pattern where women publicly engaged in issues are disproportionately subjected to harassment and disinformation.'
The Oversight Board's decisions on individual cases are binding for Meta, but its broader policy recommendations do not carry the same weight.
Meta expressed approval of the Board's decision on both cases and agreed to comply within seven days. San Martin emphasized that 'Meta and other social media platforms need more robust policies to combat the spread of deepfakes, including a much stronger approach to labeling misleading content, stricter penalties for violators, and the removal of deepfakes that target private individuals.'
Last week, the Board urged Meta not to replace professional fact-checking with the 'Community Notes' feature—a collective participation model based on the misinformation that became popular on Elon Musk's platform X. This call followed Meta's announcement of testing 'Community Notes' in 16 Spanish-speaking Latin American countries as part of a move to replace its fact-checking program.