Nvidia launches Open Agent Safety Platform to stop AI agent security breaches

On Monday, September 28, 2026, Nvidia released a new AI safety platform called the Open Agent Safety Platform, consisting of two open-source tools, OpenShell and Nvidia Sentry, designed to monitor and control AI agents in real time. The release follows several security breaches involving AI agents, including a hack of Hugging Face, which Nvidia agreed to acquire for $13 billion. Nvidia's CEO, Jensen Huang, views AI safety as an engineering problem rather than a regulatory issue. The tools are being launched with partners including Anthropic, and Nvidia claims they could have prevented the Hugging Face attack.