Nvidia has become the first major AI platform to launch with security built-in rather than added later, announcing protection from five security vendors for its agentic AI stack at GTC this week. The move comes as cybersecurity professionals increasingly view agentic AI as a top threat vector.
According to industry data, 48% of cybersecurity professionals rank agentic AI as the primary attack vector heading into 2026, while only 29% of organizations feel prepared to deploy these technologies securely. IBM's 2026 X-Force Threat Intelligence Index documented a 44% surge in attacks exploiting public-facing applications, accelerated by AI-enabled vulnerability scanning.
Nvidia CEO Jensen Huang emphasized the security imperative during his GTC keynote, stating that "agentic systems in the corporate network can access sensitive information, execute code, and communicate externally. Obviously, this can't possibly be allowed." The company has developed a unified threat model with five security vendors, including CrowdStrike, Palo Alto Networks, JFrog, Cisco, and WWT, each covering different governance layers.
The timing reflects the rapidly evolving threat landscape, where machine identities now outnumber human employees 82 to 1 in the average enterprise. The five-layer governance framework addresses agent decisions, cloud runtime protection, supply chain provenance, prompt-layer inspection, and pre-production validation, though no single vendor covers all layers comprehensively.