Nvidia says its new AI safety platform can contain rogue agents within ‘milliseconds’

Nvidia is launching a caller information level designed to incorporate and show AI agents, a determination that comes successful effect to a wave of rogue hacking incidents, arsenic reported earlier by Reuters. In an announcement connected Monday, Nvidia says its caller Open Agent Safety Platform tin quarantine agents that effort to flight their boundaries wrong "milliseconds."

The level uses Nvidia's OpenShell open-source software, which runs connected the company's Vera AI CPU. Users tin take the accusation an AI cause tin access, and OpenShell checks these restrictions earlier and during a task, according to Nvidia. It besides includes Nvidia's Sentry exertion connected a abstracted …

Read the afloat communicative astatine The Verge.