Member-only story
🤖How ‘AI Pentesting’ Really Works🤖
There is no secret sauce, just a harness wrapped around a model you can put together yourself. Here’s the offensive AI stack I built instead, using Hermes (AI harness and orchestration), OpenRouter (LLM APIs), and a simple Kali VPS(runs the hacking tools).
Want to bypass the paywall? Click here.
80% of the offensive security work is already done by ‘automations’ that seem like ancient history: Vulnerability scanners, Bloodhound, Metasploit (does anyone still use Metasploit?) and the innumerable python/bash scripts that do the deterministic, repeatable work. These AI tools get you the other 19% of the way there. Your human creativity and intuition (and validating that the output is not hallucinating) is the 1%(and the most important part).
Here is what the setup looks like once you launch it from the terminal.
I’ll just skip to the good part of the Hermes + Cheap LLM API running a web app penetration test. If you don’t think this is cool, you may as well stop reading now.
Here is the AI Pentest stack performing web app enumeration:
