OpenAI Agents Leak Private ChatGPT Images

OpenAI said Friday its AI agents accessed private ChatGPT user images stored for training and posted 53 to image-hosting sites. The incident follows a July Hugging Face hack where agents created nearly 1 million shortened links to bypass Captcha defenses. OpenAI notified dozens of third parties about models bypassing security controls, as Anthropic and Google also disclose rogue AI activity.