OpenAI says AI agents posted 53 user images online

OpenAI said Friday that its artificial intelligence agents posted 53 images from ChatGPT users to online image-hosting sites without the company's knowledge.

The company said links to the uploaded images were not publicly listed and that most of the images had been removed with help from the hosting providers. It said removal of the remaining images was underway.

OpenAI said the images came from accounts of users who had authorized use of their data to improve the company's models. The company said the data had been run through a privacy filter before use and could no longer be linked to the original user.

OpenAI did not specify whether the images depicted identifiable people or contained sensitive data.

The company also confirmed a New York Times report that its tools had accessed websites of US federal agencies. OpenAI said the tools retrieved only publicly available information.

OpenAI said research agents transmitted training and evaluation data to third-party services when they should not have done so. It said the incidents happened before it strengthened security protocols in its research environment in August after other actions by AI agents.

OpenAI said it is reviewing past activity by its AI agents, a process it said “will take months to complete.”

OpenAI chief executive Sam Altman said on X that the company had “not been as fast as we would have liked” in reviewing and disclosing the incidents. He said the company was trying to balance transparency with assessing a large volume of data.

On July 21, OpenAI disclosed that two models escaped closed test environments, reached the internet and broke into internal systems at Hugging Face, an online library for AI software. Altman said Friday that the Hugging Face incident “is still the most severe event we've seen.”