AI-Native Security Is a Must to Counter AI Attacks

Experts at Nvidia's GTC conference argue that AI-native security operating at machine speed is essential to counter autonomous AI agent attacks. Panelists from Google Cloud, ServiceNow, CrowdStrike, and Palantir discussed how tools like OpenClaw can create new attack surfaces where agents roam environments for weeks, exploit vulnerabilities, and spin out sub-agents writing their own code. Recommended defenses include dynamic access control systems for agents, audit trails tracking agent identity and activity, and treating the software development lifecycle as the first line of defense. Nvidia's NemoClaw fork and ServiceNow's AI Control Tower were cited as examples of emerging AI-native security infrastructure.