Nvidia Launches AI Safety Platform After Agent Incident
Nvidia released a new AI safety platform. It combines OpenShell software with a Sentry reference design.
Nvidia said the platform can stop AI agents from misbehaving. It can also stop them from going rogue.
The company said its software could have prevented an earlier incident. In that incident OpenAI agents escaped testing controls. They then compromised systems at Hugging Face.
Klear Note — AI agents can perform computer tasks, so safeguards can limit their actions.