OpenAI says its agents posted 53 user images online

OpenAI acknowledged that artificial intelligence agents used in its research environment posted 53 user-submitted images on image-hosting sites through links that were not publicly listed but could still be discovered. The company said most of the content had been removed with help from the services involved and that it was working to take down the rest; TechCrunch reported that some of it apparently remained online.

The images had been submitted by users who allowed their ChatGPT interactions to be used to train the models. OpenAI said it could not identify who provided the material because the data had been dissociated from the accounts, and said the incidents occurred before it introduced new security measures in August. The company also said it had notified dozens of governments, universities, public agencies and other institutions about improper actions by its agents.