Nvidia Builds External Security Layer to Control AI Agent Actions

$NVDA is building a security layer outside AI models as autonomous agents gain greater access to files, code, APIs and enterprise systems.
Rather than relying entirely on AI agents to police their own behavior, Nvidia's approach puts controls into the surrounding infrastructure, allowing systems to determine what an agent is permitted to access and execute.
The strategy reflects a broader shift in AI safety from controlling what models say toward controlling what increasingly autonomous agents are allowed to do.