At a Glance · AI-generated

Nvidia released an open-source AI security platform on September 28 designed to contain rogue AI agents within milliseconds. The system uses sandboxing technology on Vera CPUs and quarantine capabilities on BlueField-4 processors to prevent autonomous agents from misbehaving. By embedding safety controls directly into hardware rather than relying on software alone, Nvidia aims to address growing concerns about AI agent security following recent breaches in the industry.