Il board di Meta ordina la rimozione di due deepfake e
Meta Oversight Board mandates removal of two deepfakes targeting women and critiques the company's inadequate policies on synthetic content
- Incident date
- Nov 2025
- Target
- unnamed Scottish Labour councillor and an unnamed Muslim health volunteer
In November 2025, the Meta Oversight Board issued a binding decision ordering the removal of two AI-generated videos from Facebook. The board criticized Meta's existing policies as deeply inadequate, highlighting a failure to address the rapid rise of synthetic content used to harass public and private figures.
What happened
The first case involved a Scottish Labour councillor who was targeted by a deepfake video. The clip used voice cloning and face manipulation to attribute a fabricated, offensive statement to the politician regarding refugees. Despite being reported by the councillor and other users, Meta’s automated systems rejected the appeals, claiming the content did not violate rules. The board ultimately ruled that the video violated policies against hate speech by attributing predatory behavior to a protected group and mandated that it be labeled as High Risk AI. The councillor described the unauthorized use of her voice as traumatic, noting that police reports regarding the incident were unsuccessful.
The second case concerned a young Muslim woman who participated in a television campaign regarding menstrual health. Digital manipulators edited footage from her interview to depict her performing absurd actions and consuming junk food. These videos garnered tens of millions of views across various platforms, including Facebook. Meta’s automated systems closed reports on this content without human review, and the video remained online without any synthetic media labeling. The board determined this content constituted bullying and harassment.
Following these incidents, the Oversight Board issued nine recommendations, including lowering the threshold for High Risk AI labeling, demonetizing high-risk content, and expanding the definition of unwanted manipulated imagery to include deepfakes of private citizens. Meta has agreed to remove the two specific videos but has 60 days to respond to the board's broader policy recommendations, which are not binding.