An unelected board that Meta pays to police its own platform has ordered the company to remove two deepfake videos from Facebook and demanded sweeping new powers to control what millions of Americans are permitted to see online.
The Oversight Board — a body with zero democratic mandate, bankrolled by the corporation it oversees — overruled Meta's own content moderators this week and called for algorithm-driven suppression, mandatory click-through warning screens, and account penalties for users who share AI-generated content. Its individual rulings are binding on Meta; its policy recommendations are not, though the company has 60 days to respond.
The cases involve two videos the board says are AI-generated. In the first, a Scottish Labour councillor is falsely depicted saying: "Refugees are welcome here, even if they r*** our women, because white people do that too." The board found the audio was not synchronized with her facial movements — a telltale sign of fabrication. The councillor told the board the words "run contrary to everything [she] stands for" and said she reported the video to Meta and to police without success, according to TNW.
Meta initially refused to remove the video. The company said no "Trusted Partner" organization had flagged it, that it contained no banned claims about voting or the census, and that it appeared satirical with low engagement. Two users reported it, including the councillor herself, but Meta's systems never sent it for human review and both appeals failed.
The board's majority ruled the video violated Meta's hateful conduct policy by attributing "serious criminality and predatory sexual behavior" to refugees as a group. A minority disagreed, arguing the video targeted the politician's alleged views, not refugees generally. The board also said the video should have carried a "High Risk AI" label. It did not find a bullying violation because the councillor counts as a public figure.
The second case involves a young Muslim woman in Europe who participated in a menstrual health education campaign. AI-manipulated videos mocking her — including one showing her exercising absurdly and eating junk food — racked up tens of millions of views, The Guardian reported. The board ordered that video removed as a bullying violation.
Board co-chair Pamela San Martin framed the rulings as a women's issue: "From politicians to private citizens, AI-generated deepfakes are increasingly being used to harass and silence women from engaging in public discourse." The Guardian dutifully echoed that framing, calling the board Meta's "supreme court" — a telling label for a body no citizen ever voted for.
The board described Meta's existing safeguards as "consistently and fundamentally inadequate" and recommended up to nine policy changes: lowering the threshold for "high risk" AI labels beyond elections and crises, requiring click-through warning screens, demoting labeled content in algorithms, demonetizing it, penalizing repeat sharers, and publishing annual data on label application.
Here is the real stakes: the board is not just asking Meta to take down two videos. It is demanding that algorithms bury content a corporate tribunal deems "high risk," that users be forced through warning screens to view it, and that accounts be punished for sharing it. This is a censorship architecture dressed up as governance — a privately funded supreme court for speech, answerable to no voter and no law, expanding its writ with every ruling.
The question is not whether deepfakes are real or harmful. It is who gets to decide what you see, and by what authority.
