Oversight Board faults Meta’s AI deepfake rules as inadequate
The Board’s rulings spotlight how weak AI labeling and enforcement can intensify harassment, especially against women in public life.
Meta’s Oversight Board issued two new case decisions finding that Meta’s current policies and enforcement around AI‑generated videos are structurally insufficient, especially for protecting women from targeted harassment. In one case, Meta left up and failed to label a realistic AI deepfake of a Scottish politician making hateful remarks about refugees, which the Board said violated Meta’s hateful conduct rules and should have been clearly marked as AI. In the second case, Meta auto‑closed a report and appeal over an AI‑manipulated interview of a menstrual‑health volunteer used to mock her, only removing it for bullying after the Board intervened and without any AI label. The Board highlighted a broader pattern in which AI deepfakes are being used against women in public life, with Meta’s small “AI Info” banners and inconsistent labeling seen as inadequate. Meta now has 60 days to respond to the Board’s new recommendations, including giving higher review priority to content likely to be AI‑generated when users report it.
Why it matters
The decisions put direct pressure on Meta to better detect and review AI‑generated content that targets individuals, especially women who participate in public discourse. By calling out how unlabeled deepfakes and auto‑closed reports left harmful videos online, the Board is drawing a line under the risks of AI tools being used for harassment and misinformation, and setting out expectations for stronger labeling and faster review of suspected AI content across Meta’s platforms.