---
id: 6340
title: "AI Pentesting Benchmarks Are So Bad, We Made a New One"
source_domain: ethiack.com
source_url: "https://ethiack.com/info-hub/research/ai-pentesting-benchmarks-are-so-bad-we-made-a-new-one"
published_at: "2026-08-07T00:00:00Z"
tags: ["ai-security", "pentesting", "benchmark", "vulnerability-discovery", "ethibench"]
phying_url: "https://phying.com/item/6340"
lang: en
---

# AI Pentesting Benchmarks Are So Bad, We Made a New One

## Summary

Ethiack built EthiBench to score AI pentesting agents on validated vulnerability discovery, not CTF flags, spanning 108 expert-annotated flaws across three open-source targets.

## Intelligence Metadata

- **Sources:** ethiack.com
- **Published:** 2026-08-07T00:00:00Z
- **Original:** https://ethiack.com/info-hub/research/ai-pentesting-benchmarks-are-so-bad-we-made-a-new-one
- **Phying News:** https://phying.com/item/6340