Home / Technology / Nvidia Unveils AI Agent Security Tools Amidst Hacking Fears
Nvidia Unveils AI Agent Security Tools Amidst Hacking Fears
28 Sep
Summary
- Nvidia releases OpenShell security sandbox for AI agents.
- New Sentry platform monitors AI agents on Bluefield DPUs.
- Nvidia leads industry in open-source AI security initiatives.
Nvidia is rallying tech companies with new open-source AI security tools as AI agents increasingly pose threats. The company has made its OpenShell security sandbox, designed to isolate AI agent activity, generally available. This framework helps contain agents within operating system kernels to prevent unauthorized actions.
Additionally, Nvidia introduced Sentry, a software platform implemented on its Bluefield DPUs. Sentry continuously monitors AI agents, acting as an independent mechanism to quarantine any agent attempting to move beyond its designated boundaries. Both OpenShell and Sentry are part of Nvidia's new Open Agent Safety Platform.
These initiatives are part of Nvidia's broader efforts to lead the industry in AI safety, including an AI safety coalition comprising over 120 companies. The company aims to set standards from silicon to security software, addressing the evolving challenges of autonomous AI behavior and strengthening defenses against potential misuse.
