Skip to content
arXiv cs.AI · Papers

From Controlled to the Wild: Evaluation of Pentesting Agents for the Real-World

arXiv:2605.10834v3 Announce Type: replace Abstract: AI pentesting agents are increasingly credible as offensive security systems, but current benchmarks still provide limited guidance on which will perform best in real-world targets. Existing evaluation protocols assess and optimize for predefined goals such as capture