Forget the model. When it comes to cybersecurity, it’s all about the harness

Cato Networks' research shows that combining advanced large language models with an operational "harness" can autonomously perform end-to-end simulated intrusions—probing targets, discovering services (e.g., Microsoft Exchange), performing lateral movement, and achieving domain administrator privileges in some scenarios within about 40 minutes—highlighting that harness infrastructure, not just the underlying frontier model, materially increases attack effectiveness and could be repurposed by defenders or adversaries.