Overview
- Ruling published March 10 overturns Meta’s choice not to label a 2025 Israel–Iran AI video and urges a separate AI standard, at-scale C2PA provenance, stronger detection for video and audio, clearer labels, and defined penalties.
- Meta was faulted for relying on self-disclosure and limited metadata checks, with the Board noting inconsistent watermark use and advising more proactive crisis protocols rather than waiting on third‑party fact‑checkers.
- AI-generated and manipulated visuals continue to surge since February 28, including a Tehran Times satellite image traced via Google’s SynthID to a Bahrain base photo, as researchers report record viral volumes across platforms.
- Verified cases show AI‑enhanced real images shifting perceptions, such as a Kuwait incident photo and an Erbil blaze shot that detection tools tied to Google AI, with experts warning small edits can recast events.
- Platforms’ responses remain patchy: X moved to suspend revenue sharing for undisclosed AI war videos while its chatbot misjudged fakes, and separate reporting highlights US military use of Anthropic’s Claude via Palantir for rapid targeting that has intensified scrutiny of military AI.