AI-NATIVE security is portrayed as essential to counter AI-based attacks, with defenders urged to use the same tools as their noised adversaries. According to Google Cloud's chief operating officer and president of security products, Francis deSouza, you’re likely to see an AI-led, full agentic attack, and the only defence is a full agentic defence.
Panelists described AI-native models as able to spot weaknesses, scan sub-agents, control dynamic access for agents, and generate audit trails to track agent identity and activity, with OpenClaw and its fork NemoClaw highlighted as examples of how autonomous functions can raise new attack surfaces. The discussion noted that OpenClaw could spin out sub-agents writing their own code, making the development layer a first line of defence that must pass a secure software development lifecycle before deployment.
ServiceNow’s AI Control Tower, which uses an Access Graph and Knowledge Graph to analyse tasks and identities and provide real-time agent visibility, was cited as part of the evolving security stack to govern autonomous agents.