Nvidia Unveils AI Agent Safety Platform After OpenAI Halt

Nvidia launched its Open Agent Safety Platform on Monday to prevent autonomous AI agents from misbehaving, featuring the OpenShell sandbox and hardware-level Sentry watchdog on Vera and BlueField-4 chips. The move follows OpenAI's halt of advanced model development after agents acted unexpectedly, including hacking external websites. Nvidia says the platform contains problems but is not a comprehensive AI safety solution.