Meta’s quasi-independent Oversight Board ordered the tech company to remove deepfake videos targeting a Scottish politician and a young Muslim woman from Facebook, ruling on Thursday that the platform’s current safeguards against deceptive artificial intelligence are consistently and fundamentally inadequate. According to the board, a fake video showing a Labour party councillor in Scotland making inflammatory remarks about refugees and a manipulated clip mocking a female Muslim health campaign volunteer should have been swiftly taken down under existing platform policies.
Oversight Board Orders Meta to Remove Political Deepfakes
The ruling centers on two distinct pieces of AI-generated media hosted on Facebook. In the first case, a Scottish Labour councillor was falsely depicted stating that refugees are welcome even if they commit sexual assaults, because white people do so as well. According to the Oversight Board, the audio was not fully synchronized with the councillor’s facial movements, signaling clear AI manipulation. Meta initially decided the clip did not violate its content policies and declined to apply an AI label. However, the board ruled that the post violated rules against hateful conduct by attributing criminal behavior to refugees as a group, and ordered its immediate removal alongside a high-risk AI label.
Tackling Harassment and Misinformation Targeting Women
In a separate ruling, the board ordered Meta to remove an AI-manipulated video mocking a young Muslim woman in Europe who participated in a campaign promoting menstrual health education. The video showed the woman performing absurd exercises and eating junk food, accumulating tens of millions of views across social platforms. Pamela San Martin, an oversight board co-chair, noted that AI-generated deepfakes are increasingly deployed to harass and silence women in public discourse. The board concluded that the video breached Meta’s bullying and harassment policy, recommending that definitions of unwanted manipulated imagery be expanded to cover deepfakes of private individuals.
Policy Recommendations and Algorithmic Penalties
The Oversight Board issued nine policy recommendations urging Meta to overhaul how it handles deceptive media. According to the board’s blogpost, the company must expand situations where high-risk labels apply, increase penalties for accounts that repeatedly share deceptive content, and provide greater data transparency regarding AI labels. The board also recommended that Meta algorithms demote high-risk content to reduce feed visibility, and introduce warning screens requiring click-through actions before users can view AI-generated material. Meta, which established the Oversight Board in 2020 as a content referee, has been approached for comment on the binding rulings.
Did you know? The Oversight Board functions as a binding quasi-independent body designed to review content moderation decisions made by Meta across Facebook and Instagram.
Frequently Asked Questions
What did Meta’s Oversight Board rule regarding deepfakes?
The board ordered Meta to remove AI-generated deepfake videos targeting a Scottish politician and a young Muslim woman, ruling that Meta’s current safeguards against deceptive AI are fundamentally inadequate.
Why did the board say the Scottish councillor’s video violated Meta’s rules?
According to the board, the video violated hateful conduct policies by alleging criminal and predatory sexual behavior by refugees as an entire group, while also featuring unsynchronized audio indicative of AI generation.
What specific policy changes did the Oversight Board recommend?
The board recommended nine policy changes, including expanding high-risk AI labels, algorithmic demotions for labeled content, stricter penalties for repeat offenders, and click-through warning screens for viewers.