Meta Pulls AI Deepfake Videos After Oversight Board Calls

On February 6, 2025, Meta was ordered by its Oversight Board to remove two AI-generated videos from Facebook and overhaul its safeguards against synthetic media. The board, which reviews content moderation decisions, criticized Meta’s existing policies as consistently and fundamentally inadequate to handle the rapid spread of AI-manipulated content.

What happened

The rulings involved two separate instances of deepfake abuse targeting women. The first case featured a Scottish Labour councillor whose voice was cloned to depict her making inflammatory remarks linking refugees to sexual violence. Although the video was realistic, the audio was not synchronized with the councillor's facial movements. Meta had initially declined to remove the video, claiming it did not interfere with civic processes and lacked a disclosure label. The board rejected this, ruling the content violated hateful conduct rules by attributing criminal behavior to a protected group.

The second case involved a young Muslim woman in Europe who advocates for menstrual health education. Manipulated footage from a past television interview was used to create videos depicting her offering health advice while engaging in absurd behavior, such as eating junk food. These videos received tens of millions of views across social media. Meta had previously closed reports regarding this content, only removing it after the Oversight Board intervened.

The board issued nine policy recommendations to address these failures, including the mandatory use of warning screens for AI content, demoting high-risk AI in feeds, and increasing penalties for repeat offenders. The board specifically noted that women in public life are disproportionately targeted by such deepfakes. Meta has committed to removing the two videos within seven days and has 60 days to respond to the board's broader policy proposals.