Nvidia Launches Agent Safety Platform Amid Record Share Repurchase Authorization

Nvidia has introduced the NVIDIA Open Agent Safety Platform, an open software stack and reference design aimed at keeping AI agents within defined limits from testing to production.

The platform consists of two main components: OpenShell, an open-source secure runtime that enforces policy as work happens, and Sentry, a hardware watchdog that runs on BlueField-4 data processing units.

Nvidia's CEO Jensen Huang framed the launch as a critical step in realizing AI's potential for society, stating 'Safety and security require full-stack engineering.'

The platform has garnered support from over 100 companies across various industries, including Anthropic, Microsoft, Hugging Face, Salesforce, SAP, Scale AI, JPMorgan Chase, CrowdStrike, and robotics firms such as Figure.