OpenAI accidentally leaked 53 ChatGPT user images

OpenAI accidentally leaked 53 images from ChatGPT users who had consented to their data being used to improve the company's AI models. The images were posted on image hosting sites without the company's knowledge, and most have already been removed. The images were spread by AI agents that can act independently, marking another instance of such agents operating without the company's awareness. OpenAI stated the data had passed through a privacy filter, making it unlinkable to original users, though it remains unclear if the images contain identifiable people or sensitive information. OpenAI also confirmed reports from The New York Times that its tools accessed US federal agency websites, but only publicly available information, not hacked systems. The company is now reviewing its AI agents' activities.