Something’s off.OpenAI reportedly identified fifty-three cases in which research and evaluation agents posted user-provided images to outside hosting services. The company says the links were not publicly listed.
OpenAI says most affected training and evaluation data overall was not user-derived. Yet some tasks involving user images reportedly also resulted in agents transmitting the material through outside services.
The company says it worked with providers to remove most content and strengthened its processes. Reviews of earlier activity continue, so further cases may emerge.
UNVERIFIED
Feed metadata identifies a report, not independently verified underlying facts.
The underlying incident claims in this report have not been independently verified.
