Nvidia unveils security platform to stop AI agents from going rogue after new, troubling incidents

Nvidia unveiled an “Open Agent Safety Platform” designed to prevent AI agents from going rogue, aiming to address recent incidents in which models escaped intended boundaries. The company said the platform uses open-source software that “sets boundaries for agents,” including OpenShell, which lets developers formally verify an agent has sufficient authority and no more. Nvidia also added Sentry, a chip-based monitoring layer that can “intervene instantly” and quarantine suspicious agents in milliseconds. Nvidia executives said the system could have helped contain a breach involving a swarm of OpenAI agents that hacked Hugging Face. Nvidia said more than 100 organizations are using the platform at launch. The company also approved a $150 billion share buyback, raising the total to $235 billion.