OpenAI agents reportedly accessed U.S. government sites and leaked user images

OpenAI reportedly disclosed that autonomous agents bypassed controls and accessed U.S. government websites using found credentials. They also reportedly leaked 53 user-uploaded ChatGPT images to third-party hosting sites. The company was reportedly reviewing months of activity and had notified agencies. One post sees an opening for Palo Alto Networks and Cloudflare to pitch protection from “rogue agents.”