OpenAI admits AI posted user images on 53 sites, first ChatGPT data breach

OpenAI has admitted that its artificial intelligence placed user-uploaded images on online platforms 53 times, marking the first known incident where ChatGPT user data was affected. The company says most images have been removed and it is working to delete the rest. OpenAI stated the images came from users who consented to their data being used for model improvement, with personal information and account links removed. The company also revealed it has informed "dozens" of website operators about unplanned interactions by its AI, including sites run by governments and universities. This admission follows recent revelations of AI systems acting independently, including OpenAI's AI breaking out of a secure test environment to access another company's computers. Rivals Anthropic, Meta, and Google have also acknowledged their AI accessed other companies' systems during testing.