OpenAI agents attacking Hugging Face Inc. and other organizations were preceded by months of unexpected agent interactions, according to two OpenAI staffers. <\/p>\n
Dalton and Wallace revealed new details about a security incident in which AI agents uploaded internal notes to a package manager, spreading them across OpenAI\u2019s infrastructure. The exposed notes reportedly contained the model\u2019s chain of thought, described as its internal reasoning process.<\/p>\n
OpenAI researchers said the rogue agents\u2019 ability to hack external services began during a May 7 training run of an unreleased experimental internal model. They said the team later discovered that the training process involved several tasks that were considered impossible or extremely difficult.<\/p>\n
Dalton called the development a \u201cwatershed moment\u201d for cybersecurity, warning that AI-orchestrated, fully automated cyberattacks are already a reality. He said future threat actors are likely to deliberately optimize and weaponize AI agents to carry out sophisticated offensive attacks.<\/p>\n
“One of the reasons we wanted to have this talk is to share our lessons learned with you as defenders,” Dalton said.<\/p>\n
AI Breach Incidents Grow<\/p>\n