Meta’s strategies for detecting deepfakes are falling short, especially in the fast-paced spread of misinformation during conflicts like the ongoing situation in Iran. This critique comes from the Meta Oversight Board, an advisory body that influences the company’s content moderation. The board is urging Meta to revamp how it identifies and labels AI-generated content on its platforms, including Facebook, Instagram, and Threads.
This push for change follows an inquiry into a fabricated AI video depicting alleged destruction in Israel, which circulated on Meta’s platforms last year. The board emphasizes that its recommendations are crucial now, given the recent “massive military escalations” in the Middle East. They assert that ensuring access to accurate and trustworthy information is crucial for public safety, especially with the rising potential for AI tools to disseminate false information.
The Meta Oversight Board noted, “Our findings reveal that Meta’s current approach to labeling AI content relies too heavily on self-reporting and elevated scrutiny, which is inadequate for today’s digital landscape.” They pointed out that the problematic content often originates from platforms like TikTok before spreading to Facebook, Instagram, and X.
To address these issues, the board has suggested that Meta strengthen its misinformation policies to specifically tackle deceptive deepfakes and introduce a distinct community standard for AI-generated materials. They also recommend that Meta enhance its AI detection capabilities, clarify penalties for policy breaches, and expand efforts to label AI content. Part of this expansion involves more consistently applying “High-Risk AI” labels to synthetic media and enhancing the adoption of Content Credentials (C2PA) to ensure that information about AI-generated content is readily visible and accessible to users.
