{"id":"b87193f7-507d-4594-81be-845ca1c0a4c2","url":"https://www.axios.com/2026/09/28/nvidia-ai-agent-safety","domain":"axios.com","outlet":{"name":"Axios","country":"US","cred":4,"lean":0,"curated":true},"title":"Nvidia says new tool can contain rogue AI agents in \"milliseconds\"","plainHeadline":"Nvidia says new tool can quarantine rogue AI agents in milliseconds","rewrite":{"model":null,"at":null},"summary":["Nvidia introduced the Nvidia Open Agent Safety Platform with OpenShell and Sentry.","The system traces all actions by agents running on Vera CPUs and quarantines them quickly.","Nvidia argues guardrails enable safe AI deployment while accelerating development.","OpenAI, Anthropic, and researchers report numerous incidents with frontier models.","The rollout signals growing demand for chips, data centers, and power to support AI safety."],"topics":["AI","Cybersecurity"],"lang":"eng","publishedAt":"2026-09-28T21:20:42.000Z","ingestedAt":"2026-09-28T21:24:41.509Z","story":"https://brief.concise.news/story/665f64c5-e8b1-4132-8c9c-e7f2a8c83d51","storyId":"665f64c5-e8b1-4132-8c9c-e7f2a8c83d51","image":"https://images.axios.com/59L6uT7Jl_0zxvuTa0W1t19-__g=/0x58:3592x2078/1366x768/2026/09/28/1790627586143.jpeg","author":"Nathan Bomey","dek":"Nvidia is deploying a new tool that it says can be used to prevent and contain rogue and potentially dangerous AI agents."}