Oversight Board blasts 'inadequate' Meta safeguards for AI
Meta removed AI-generated deepfakes targeting a Scottish politician and a European campaigner following criticism from its Oversight Board regarding.
- Incident date
- Feb 2024
- Target
- A Labour Party councilor in Scotland and a Muslim woman campaigner in Europe
In February 2024, Meta removed two AI-generated videos targeting a Scottish Labour Party councilor and a Muslim woman campaigner after its Oversight Board condemned the company's existing content moderation as fundamentally inadequate. The board, which is funded by Meta but operates independently, ruled that these specific instances of deepfakes undermined company policies regarding misinformation, harassment, and hate speech.
What happened
The incident involved two distinct fabricated videos designed to discredit public figures. The first video depicted a Labour Party councilor in Scotland allegedly making incendiary comments regarding refugees. The fabrication was identified by clear technical flaws, most notably audio that failed to synchronize with the subject's facial movements. There was no public record of the politician ever making such statements.
The second video targeted a Muslim woman campaigner in Europe who serves as a health advocate. The content was designed to undermine her professional credibility by falsely depicting her consuming junk food while offering advice on mindful eating, accompanied by footage of her performing absurd exercises.
Oversight Board Co-chair Pamela San Martin noted that these incidents reflect a broader, troubling pattern where women engaged in public discourse are disproportionately targeted by deepfakes intended to harass and silence them. While Meta agreed to comply with the board's binding decision to remove these specific videos within seven days, the board continues to push for more robust systemic changes. Their recommendations include implementing stricter penalties for repeat offenders, enforcing more rigorous labeling of deceptive AI content, and specifically prioritizing the removal of deepfakes that harass private individuals. The board also explicitly cautioned Meta against replacing professional fact-checking programs with crowdsourced models, emphasizing the need for stronger institutional safeguards against the rising tide of AI-generated misinformation.