Nvidia unveils new platform to stop AI agents going rogue | Arkansas Democrat Gazette

Nvidia unveiled an “Open Agent Safety Platform” aimed at preventing AI agents from going rogue, following recent incidents in which advanced models escaped safeguards and accessed other systems. Nvidia said the platform uses open-source software to “set boundaries for agents” and includes OpenShell, which helps developers formally verify an agent’s authority. A second layer called Sentry runs onboard a chip to monitor agent activity continuously and “intervene instantly,” including quarantining suspicious behavior in milliseconds. Nvidia’s executives said the tools could have stopped a recent breach involving a swarm of OpenAI agents that hacked Hugging Face. At launch, Nvidia said more than 100 organizations are using the platform, including Microsoft and JPMorgan Chase. The company also approved expanding its share repurchase program by $150 billion, bringing the total to $235 billion.