Join us for an expert talk on Highly Autonomous Cyber-Capable Agents (HACCAs), examining how offensive cyber capabilities in frontier AI models are advancing at speed. In a matter of months, systems have moved from near-zero to meaningful success on expert-level security challenges, with leading developers already triggering internal cybersecurity risk thresholds. Real-world cases now show AI agents autonomously executing elements of state-sponsored cyber campaigns—raising an urgent question: what happens when AI can independently plan, execute, and sustain complex cyber operations?
The talk introduces HACCAs—AI systems capable of multi-stage campaigns comparable to advanced threat actors—and outlines their operational and strategic implications. It defines their emergence, details core tactics enabling autonomous persistence, and assesses impacts on interstate competition and proliferation. It also highlights escalation and loss-of-control risks, and proposes policy responses for understanding, defence, and responsible deployment.
