Facebook’s independent oversight board has ruled the platform was wrong to leave up AI-generated videos falsely depicting a Scottish Labour councillor and a Muslim health campaigner, warning that Meta’s protections against deepfakes are “consistently and fundamentally inadequate.”
Meta’s independent oversight board, often described as the company’s “supreme court,” has ordered the removal of deepfake videos targeting a UK politician and a young Muslim campaigner from Facebook, while calling on the tech giant to strengthen its approach to AI-generated fake content. The board ruled that a fabricated video showing a Labour councillor in Scotland making inflammatory remarks about refugees should never have been permitted to remain online. In a separate case, it found that an AI-generated video falsely portraying a Muslim health campaign volunteer performing absurd exercises and eating junk food should also have been taken down.
Fake video misrepresented Scottish councillor’s views
In the first case, an unnamed Labour councillor in Scotland was falsely depicted in the video saying: “Refugees are welcome here, even if they rape our women, because white people do that too.” The oversight board said the clip appeared to be AI-generated, noting that the audio was not fully synchronised with the councillor’s facial movements. Despite the issue being raised directly with the company, Meta had determined that the video did not breach its content policies and did not warrant an AI-generated content label.
In a blogpost published on Thursday, the board said the video should have been removed for violating Meta’s hateful conduct rules, as it attributed criminal and predatory sexual behaviour to refugees as a group rather than to specific individuals. It added that the post should also have carried a “high risk AI” label under Meta’s own policies.
Board says Meta’s deepfake policies fall short
The oversight board said its findings pointed to a wider failure by Meta to keep pace with the growing threat posed by AI-generated deepfakes. “The majority [of the board] finds that Meta needs more robust policies on deepfakes, including expanding the situations when ‘high risk’ labels can be applied, more measures to reduce the spread of deceptive AI content, increasing the penalties for accounts that repeatedly share it, and more transparency on data around when AI labels are applied,” it said.
Pamela San Martin, co-chair of the oversight board, said the two cases reflected a broader trend of AI being weaponised against women in public life. “From politicians to private citizens, AI-generated deepfakes are increasingly being used to harass and silence women from engaging in public discourse,” she 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.”
Nine policy changes recommended
As part of its ruling, the board issued nine policy recommendations for Meta to adopt. These include ensuring that algorithms demote content labelled “high risk,” reducing the likelihood of such posts appearing in users’ feeds, and introducing measures such as a warning screen that requires users to click through before viewing AI-generated content.
Separate ruling over video targeting Muslim health campaigner
In a second case, the board ordered Meta to remove an AI-generated video from Facebook that mocked a young Muslim woman in Europe involved in a campaign to improve menstrual health education and reduce stigma around women’s health issues within ethnic minority communities. According to the board, AI-manipulated videos and images targeting the woman had collectively amassed tens of millions of views online, including across Meta’s own platforms.
The board identified one video in particular – showing the woman performing exaggerated exercises and eating junk food – as a clear breach of Meta’s bullying and harassment policy, and recommended its removal. It made three further policy recommendations in relation to this case, including a call for Meta to broaden its definition of “unwanted manipulated imagery” to explicitly cover deepfakes that show private individuals saying or doing things they never actually said or did.
About the oversight board
The oversight board was established by Meta in 2020 to act as an independent referee for content decisions across its platforms, with the authority to issue binding rulings on individual cases and non-binding policy recommendations to the company. Meta has been approached for comment on the board’s latest findings.
