Meta’s Oversight Board has criticised the company’s safeguards against AI-generated deepfakes as inadequate, ordering the removal of two fabricated videos that targeted women and calling for stronger measures to limit the spread of deceptive content.
The board ruled that one video, posted on Facebook, falsely portrayed a Scottish Labour councillor making inflammatory comments about refugees. The video showed signs of artificial manipulation, including audio that did not properly match the woman’s facial movements.
The second video targeted a Muslim campaigner in Europe and falsely depicted her engaging in behaviour intended to undermine her work as a health advocate. The board said both cases demonstrated how AI-generated content can be used to harass individuals and damage their reputations.
The Oversight Board said Meta’s existing policies were not sufficient to deal with the growing volume and sophistication of AI-generated material appearing on its platforms, including Facebook and Instagram.
The board also highlighted what it described as a broader pattern in which women involved in politics, journalism and activism are targeted with fabricated content intended to discredit or silence them.
Meta had initially decided not to remove the videos, but the board overturned those decisions. Meta said it accepted the rulings and would remove the content within seven days.
The board has called for stronger policies covering deepfakes, including wider use of prominent warnings for high-risk AI-generated material and measures to reduce the ability of deceptive content to spread widely.
It also recommended tougher penalties for accounts that repeatedly share harmful deepfakes, along with greater transparency about when and how AI-generated content is labelled.
The board said Meta should also strengthen its policies covering manipulated images involving private individuals, where victims may have limited ability to respond publicly or seek effective remedies.
The recommendations come as social media platforms face growing pressure to address the misuse of generative artificial intelligence. Increasingly realistic videos and audio can make fabricated statements or events appear genuine, creating challenges for users, regulators and technology companies.
The Oversight Board’s individual case decisions are binding on Meta, although its broader policy recommendations are not. The company will therefore need to decide how to implement the wider changes proposed by the board.
The board has also recently raised concerns about Meta’s move towards Community Notes, a crowdsourced approach to correcting false information, arguing that professional fact-checking should not simply be replaced by community-based systems.
The latest cases add to wider concerns about how platforms can distinguish legitimate AI-generated content from material designed to deceive, harass or manipulate audiences as the technology becomes increasingly accessible.
Leave a Reply