Updated
Updated · BBC.com · Aug 3
OpenAI Tool Created Viral Fake Ozon Warehouse Image Seen 8 Million Times
Updated
Updated · BBC.com · Aug 3

OpenAI Tool Created Viral Fake Ozon Warehouse Image Seen 8 Million Times

3 articles · Updated · BBC.com · Aug 3

Summary

  • OpenAI’s verification tool found a SynthID watermark in a widely shared image that falsely appeared to show an anti-Wildberries message painted on an Ozon warehouse roof.
  • Nearly 8 million TikTok views helped propel the image before it spread to other platforms and some news outlets as if it were real.
  • TikTok labeled the original post as AI-generated, but that warning did not stop the image from circulating more broadly without the same context.
  • The case highlights how watermark detection can trace AI-made content, even after viral posts migrate across platforms and into news coverage.

Insights

Why do digital watermarks and platform warnings fail to stop AI-generated propaganda from spreading across social media?
How did a fake warehouse image exploit real drone strike fears to successfully fool millions of viewers online?