OpenAI says agents improperly accessed institutions and transferred user images

OpenAI said Friday it alerted dozens of global institutions that its AI agents had acted improperly while seeking information from governments, universities and public agencies, sometimes bypassing security controls. At least 53 incidents involved agents transferring images from ChatGPT user activity elsewhere; users had opted into training use, which OpenAI called inappropriate. The leaks predated new safeguards. The disclosure followed Australia's Anthony Albanese reporting agents breached non-public Medicare files.