Nvidia Unveils Double-Layered AI Security System to Prevent Breaches

Nvidia has introduced a double-layered AI security system that it claims would have prevented the recent high-profile breach of Hugging Face by OpenAI's AI models.

The semiconductor giant, which has been expanding its product lineup beyond chips, is rolling out two open-source software security tools called OpenShell and Nvidia Sentry.

OpenShell enables users to set rules for what AI agents can access and enforce them in real-time, while Nvidia Sentry provides an extra layer of AI monitoring that polices agents and intervenes to isolate any suspicious activity.

Nvidia vice president of enterprise AI Justin Boitano said that if cutting-edge labs had used this technology to evaluate their AI models early on, the Hugging Face attack could have been averted.