The PEN testing local LLMs I’ve tried are okay. If you know any PEN testing yourself it’s great for automation of the slog tasks (password testing, sql injection, etc…). After that though you are WAY better off using your own experience than going further with AI as it will confidently tell you X is fine or not fine when in reality it’s flipped.
Yet again, AI is great when used as a tool with something you ALREADY know, horrible to learn anything or build something you don’t understand as you are in 100% code debt.
Completely agree, it’s why I don’t mess around with the pen-testing AI models myself, I only know enough about it to be dangerous but not enough to actually fact check what they’re doing, it’s just not useful for me without my own confidence and understanding. It’s just not my field. I stick to my lane and treat my AI like particularly stupid junior employees and direct them accordingly and validate what they’re giving me and it works pretty alright as long as I maintain a narrow enough scope and don’t rush things the way a lot of AI slop pushers tend to do. It still takes time, and thought, it’s a tool, sometimes a decent one, sometimes a terrible one, but never more than a tool.
The PEN testing local LLMs I’ve tried are okay. If you know any PEN testing yourself it’s great for automation of the slog tasks (password testing, sql injection, etc…). After that though you are WAY better off using your own experience than going further with AI as it will confidently tell you X is fine or not fine when in reality it’s flipped.
Yet again, AI is great when used as a tool with something you ALREADY know, horrible to learn anything or build something you don’t understand as you are in 100% code debt.
Completely agree, it’s why I don’t mess around with the pen-testing AI models myself, I only know enough about it to be dangerous but not enough to actually fact check what they’re doing, it’s just not useful for me without my own confidence and understanding. It’s just not my field. I stick to my lane and treat my AI like particularly stupid junior employees and direct them accordingly and validate what they’re giving me and it works pretty alright as long as I maintain a narrow enough scope and don’t rush things the way a lot of AI slop pushers tend to do. It still takes time, and thought, it’s a tool, sometimes a decent one, sometimes a terrible one, but never more than a tool.