PentestBench Launches: 17-App Corpus for Testing LLM Web Vulnerability Hunting

hunarbatra · x · 2026-11-18

ananayarora's team launched PentestBench, an open-source benchmark with a 17-app corpus for testing how well different model and harness configurations find, chain, and verify vulnerabilities in modern web apps. They found web vuln hunting — often considered "easy" — still has a long way to go for AI.

Related event: PentestBench: Open Benchmark Tests LLM Web Exploitation Skills(2 posts)→

Original post →

More from coding & agent

coding & agent channel →