Safeguarding the Future in the AI Era: Q&A with Sella Nevo

RAND Corporation

The RAND Center on AI, Security, and Technology (CAST) is warning that autonomous artificial intelligence models capable of executing end-to-end cyber operations without human intervention could soon be diffused to global cybercriminals. This imminent shift threatens to eliminate existing technical bottlenecks for cybercrime, enabling unprecedented scales of automated network compromise.

To address these emerging threats, CAST director Sella Nevo is leading critical policy and technical initiatives to secure frontier AI model weights and establish robust physical and digital data center standards capable of resisting sophisticated nation-state attacks. Collaborative evaluations under the Canary initiative with METR reveal that advanced models are also approaching the capability to autonomously synthesize pathogens. Consequently, the center is prioritizing the integration of agent-based testing and open-source model evaluations to help international governments and private industry partners secure critical digital and biological systems before offensive capabilities permanently outpace defensive safeguards.

Comment

The transition of artificial intelligence to an autonomous operational agent, as evaluated by METR under the Canary initiative, fundamentally alters network defence. Traditional threat signatures and patch-deployment cycles are rendered obsolete when autonomous models execute multi-stage network compromises in real time. This shift compresses the defensive decision window to milliseconds. Consequently, the primary vulnerability lies in the systemic latency of human-centric command structures confronting machine-speed offensive operations.

To read in details click here