Nvidia CEO Unveils Israeli-Developed AI Security Tech

Jensen Huang, CEO of chip giant Nvidia, has introduced a new security architecture with Israeli core development at the GTC conference in Berlin. This technology aims to address the critical challenge of preventing AI agents, which can perform tasks autonomously, from acting against instructions and endangering organizational data. As companies grant these agents access to internal systems, sensitive documents, and the ability to act on behalf of employees, the risks of errors, data leaks, and misuse of permissions increase.

Nvidia's solution allows organizations to predefine agent capabilities, monitor their activities, and automatically block them if they attempt unauthorized actions. The new system integrates Nvidia's previously introduced open-source software, OpenShell, with hardware-level security mechanisms. Central to this hardware-based protection is the BlueField-4 chip, developed at Nvidia's Israeli R&D center, building on technology from the Israeli company Mellanox, acquired by Nvidia in 2020.

A separate system called Sentry runs on the BlueField-4 chip, independently monitoring agent activity and enforcing security policies. Nvidia claims this system can detect anomalies and isolate agents within milliseconds, without relying solely on the agent's operating software. The company is also promoting this solution in collaboration with Anthropic, the developer of Claude.

Huang explained that the goal is to enable companies worldwide to adopt and autonomously operate AI agents within their organizations, tailoring access permissions to each task and user. He emphasized that the dynamic nature of AI agent operations requires dynamic permission management, policy enforcement, and monitoring, describing OpenShell as a revolutionary technology. This development is significant for Israel's cybersecurity industry, as it positions security as an integral part of AI infrastructure rather than an add-on.