Meta urged to boost oversight of fake AI videos

Meta’s internal Oversight Board has issued a stern rebuke to the social media conglomerate, demanding comprehensive reforms to address the rampant spread of AI-generated deceptive content across its platforms. The 21-member supervisory body specifically criticized Meta’s inadequate handling of a fabricated video depicting extensive damage in Haifa, Israel, allegedly caused by Iranian forces—content created entirely through artificial intelligence tools.

The board emphasized that Meta’s current reliance on user self-disclosure for AI-generated content identification has proven fundamentally insufficient, particularly during military conflicts when misinformation spreads rapidly. This systemic failure has severely undermined public capacity to distinguish factual reporting from fabrication, potentially eroding trust in all digital information sources.

Established in 2020 as a semi-independent content moderation oversight mechanism, the board noted that despite frequent disagreements with Meta’s rulings, the company has continued relaxing its content policing approaches. The Haifa video case exemplifies persistent inefficiencies in Meta’s conflict response protocols, where content remains unlabeled until user complaints trigger review processes.

The controversial video originated from a Philippines-based Facebook account posing as a news source in June, among numerous AI-fabricated videos that accumulated over 100 million views during recent Middle East tensions. Despite clear artificial creation and multiple user reports, Meta initially refused labeling or removal, claiming the content didn’t directly risk imminent physical harm—a standard the board deemed unacceptably high for conflict-related material.

Only after direct appeal to the Oversight Board did Meta engage with the concerns, ultimately agreeing to label the specific video within seven days while committing to apply similar treatment to identical content in equivalent contexts. The board insists Meta must proactively label deceptive AI content more frequently through robust systems capable of addressing the scale and velocity of synthetic media proliferation, especially during crises.