Overview
- Fact checks published on Sept. 2–3 reviewed multiple widely shared clips and concluded the examined videos do not show the real Gyirong Port flash flood.
- Automated tools including Google SynthID, Hive Moderation, WasitAI and Sightengine flagged the clips with high confidence and found an invisible SynthID watermark in at least one video.
- Human visual forensics pointed to clear artifacts such as people merging, objects appearing or disappearing, jumping or distorted timestamps, and skylines that do not match the hilly Nepal–Tibet border.
- Investigators also traced one viral clip to an Instagram post from July 11 that predates the disaster, showing how older or recycled footage can be repurposed and mislabeled as new.
- Fact‑checkers urge users and journalists to verify provenance, check timestamps and source accounts, and use published toolkits and detectors because synthetic clips can mislead the public and hamper rescue communications during real disasters.