Salvato in:
| Autori principali: | , , , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2603.11214 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866917349570904064 |
|---|---|
| author | Folkerts, Linus Payne, Will Inman, Simon Giavridis, Philippos Skinner, Joe Deverett, Sam Aung, James Zorer, Ekin Schmatz, Michael Ghanem, Mahmoud Wilkinson, John Steer, Alan Hong, Vy Wang, Jessica |
| author_facet | Folkerts, Linus Payne, Will Inman, Simon Giavridis, Philippos Skinner, Joe Deverett, Sam Aung, James Zorer, Ekin Schmatz, Michael Ghanem, Mahmoud Wilkinson, John Steer, Alan Hong, Vy Wang, Jessica |
| contents | We evaluate the autonomous cyber-attack capabilities of frontier AI models on two purpose-built cyber ranges-a 32-step corporate network attack and a 7-step industrial control system attack-that require chaining heterogeneous capabilities across extended action sequences. By comparing seven models released over an eighteen-month period (August 2024 to February 2026) at varying inference-time compute budgets, we observe two capability trends. First, model performance scales log-linearly with inference-time compute, with no observed plateau-increasing from 10M to 100M tokens yields gains of up to 59%, requiring no specific technical sophistication from the operator. Second, each successive model generation outperforms its predecessor at fixed token budgets: on the corporate network range, average steps completed at 10M tokens rose from 1.7 (GPT-4o, August 2024) to 9.8 (Opus 4.6, February 2026). The best single run completed 22 of 32 steps, corresponding to roughly 6 of the estimated 14 hours a human expert would need. On the industrial control system range, performance remains limited, though the most recent models are the first to reliably complete steps, averaging 1.2-1.4 of 7 (max 3). |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_11214 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Measuring AI Agents' Progress on Multi-Step Cyber Attack Scenarios Folkerts, Linus Payne, Will Inman, Simon Giavridis, Philippos Skinner, Joe Deverett, Sam Aung, James Zorer, Ekin Schmatz, Michael Ghanem, Mahmoud Wilkinson, John Steer, Alan Hong, Vy Wang, Jessica Artificial Intelligence Machine Learning We evaluate the autonomous cyber-attack capabilities of frontier AI models on two purpose-built cyber ranges-a 32-step corporate network attack and a 7-step industrial control system attack-that require chaining heterogeneous capabilities across extended action sequences. By comparing seven models released over an eighteen-month period (August 2024 to February 2026) at varying inference-time compute budgets, we observe two capability trends. First, model performance scales log-linearly with inference-time compute, with no observed plateau-increasing from 10M to 100M tokens yields gains of up to 59%, requiring no specific technical sophistication from the operator. Second, each successive model generation outperforms its predecessor at fixed token budgets: on the corporate network range, average steps completed at 10M tokens rose from 1.7 (GPT-4o, August 2024) to 9.8 (Opus 4.6, February 2026). The best single run completed 22 of 32 steps, corresponding to roughly 6 of the estimated 14 hours a human expert would need. On the industrial control system range, performance remains limited, though the most recent models are the first to reliably complete steps, averaging 1.2-1.4 of 7 (max 3). |
| title | Measuring AI Agents' Progress on Multi-Step Cyber Attack Scenarios |
| topic | Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2603.11214 |