测量人工智能代理在多步骤网络攻击场景中的进展

The study evaluates the autonomous cyber-attack capabilities of cutting-edge AI models on two purpose-built cyber ranges: a 32-step corporate network attack and a 7-step industrial control system attack, which require chaining heterogeneous capabilities across extended action sequences. By comparing seven models released over an eighteen-month period (from August 2024 to February 2026) at varying inference-time compute budgets, two capability trends are observed. First, model performance scales log-linearly with inference-time compute, with no observed plateau; increasing from 10M to 100M tokens yields performance gains of up to 59%, requiring no specific technical sophistication from the operator. Second, each successive model generation outperforms its predecessor at fixed time budgets.