Meta's oversight board has ordered the company to remove AI-generated deepfake videos targeting a UK politician and a Muslim woman, criticizing the tech giant's safeguards as "consistently and fundamentally inadequate." The board recommended policy changes to better combat deceptive AI content.
The oversight board's rulings highlight the growing challenge social media platforms face in moderating AI-generated deepfakes, particularly those targeting women and public figures, potentially impacting public discourse and individual reputations.
Meta has been ordered by its oversight board to remove deepfake videos targeting a UK politician and a Muslim woman, with the board criticizing the company's safeguards against AI-generated fake imagery as "consistently and fundamentally inadequate." The board, which acts as a referee for content on Meta's platforms, ruled that a video falsely showing a Labour party councillor in Scotland making inflammatory comments about refugees should have been removed. It also found that an AI-generated video of a Muslim campaign volunteer, falsely depicting her offering health advice while engaging in absurd activities, should have been taken down.
Pamela San Martin, an oversight board co-chair, stated that AI-generated deepfakes are increasingly used to harass and silence women in public discourse, noting a pattern of misinformation disproportionately targeting women who engage publicly on issues. The board recommended nine policy changes for Meta, including expanding the application of "high risk" AI labels, reducing the spread of deceptive AI content, increasing penalties for repeat offenders, and enhancing transparency around AI labels. Specific recommendations include algorithmic demotion of high-risk content and the introduction of click-through warning screens for AI-generated material.
In a separate ruling, the board also mandated the removal of an AI-generated video mocking a young Muslim woman in Europe involved in a campaign for menstrual health education. This video, along with others targeting the woman, had garnered tens of millions of views. The board singled out one clip showing the woman exercising and eating junk food as a violation of Meta's bullying and harassment policy, recommending its removal and a change to the definition of "unwanted manipulated imagery" to encompass deepfakes of private individuals.