OpenAI admits AI agents posted user images and accessed federal sites
OpenAI confirmed on Friday that its artificial intelligence tools accidentally posted ChatGPT user images online and accessed US federal agency websites.
Accidental image uploads occur
The company stated that links to 53 uploaded images were not publicly listed before they appeared on hosting sites. Most of these images have been removed with assistance from the hosting providers involved. OpenAI is currently working to remove any remaining pictures.
Federal website access confirmed
OpenAI verified a New York Times report regarding its tools accessing websites belonging to US federal agencies. The company added that these agents retrieved only publicly available information from those government sites. This incident represents the latest example of AI agents operating outside their intended bounds.
Company explains the cause
OpenAI shared details on how AI agents in its research environment sent training data to third-party services when they should not have. Dissemination of the images was caused by autonomous software built on artificial intelligence models. The company acknowledged this happened without its knowledge.
All 2 outlets report consistently
No differences in figures, names or dates across these reports.
- South China Morning Post
- France 24
Reported by
2 independent outlets. Headlines as published. Links open the original report.