Meta oversight board orders two deepfake removals, finds safeguards inadequate

Board recommends click-through warnings and expanded ‘high risk’ AI labels

Meta’s oversight board on Thursday ordered the company to take down two deepfake videos from Facebook and called for a series of policy changes aimed at how the platform labels, distributes, and penalizes synthetic content. In a blog post, the board said Meta’s existing safeguards are insufficient to address what it described as a rapidly expanding category of deceptive imagery.

The first ruling concerned an AI-generated video falsely depicting an unnamed Labour party councillor in Scotland. The clip showed the councillor apparently saying: “Refugees are welcome here, even if they rape our women, because white people do that too.” The oversight board, a quasi-independent body whose decisions are binding, said the video of the councillor appeared to be AI-generated, as indicated by the audio not being fully synchronised to the councillor’s facial movements. Even when the board raised the video directly with Meta, the company concluded the clip did not violate its content policies and did not merit an AI label.

The board ruled the post should have been removed because it violated Meta’s rules on hateful conduct by alleging criminal and predatory sexual behaviour by refugees as an entire group rather than as individuals. The video also should have received a “high risk AI” label, the board said.

In a separate ruling, the board ordered Meta to remove AI-manipulated videos and images mocking a young Muslim woman in Europe who was volunteering for a campaign to improve menstrual health education and reduce stigma around health conditions for women and girls from ethnic minority backgrounds. The board said the manipulated content received tens of millions of views online, including on Meta platforms. It singled out one video — showing the woman exercising absurdly and eating junk food — as a breach of the company’s bullying and harassment policy.

Oversight board co-chair Pamela San Martin framed the rulings as part of a broader pattern. “From politicians to private citizens, AI-generated deepfakes are increasingly being used to harass and silence women from engaging in public discourse,” San Martin said. “These cases demonstrate a broader, troubling pattern in which women who engage publicly on issues are disproportionately subjected to harassment and misinformation. Meta and other social media platforms need more robust policies to address the proliferation of deepfakes.”

The board issued nine policy recommendations. It called on Meta to expand the situations in which “high risk” labels can be applied to AI-generated content; to implement additional measures to reduce the spread of deceptive AI material; to increase penalties for accounts that repeatedly share such content; and to provide more transparency around when AI labels are applied.

The board also recommended that Meta’s algorithms actively demote content labeled “high risk,” making such posts less likely to appear in users’ feeds. It proposed making AI-generated content harder to view by introducing a warning screen that requires a click-through before the content displays.

For the second case, the board issued three additional recommendations. It said Meta should change its definition of “unwanted manipulated imagery” to cover deepfakes depicting a private individual saying or doing things they did not say or do.

The oversight board was set up by Meta in 2020 to serve as a referee for content on its platforms, and its decisions on individual cases are binding on the company. Meta has been approached for comment on the rulings.