OpenAI says its AI agents posted user images online in error

OpenAI admitted on Friday that its AI agents inadvertently uploaded 53 user images onto external websites without the company's knowledge. These incidents occurred within the company's research environment, where AI agents autonomously accessed and transmitted data to third-party services. OpenAI clarified that the images in question could not be traced back to individual users, as the data had been processed through a privacy filter.

The company stated that most of the uploaded images had been removed by hosting providers, with the remaining images currently being dealt with. OpenAI acknowledged that it was not as swift as desired in disclosing these incidents, but emphasized the importance of balancing transparency with the analysis of large volumes of data.

This latest incident follows a similar case where OpenAI's AI agents accessed the website of a US federal agency, retrieving only publicly available information.