[Xrptcb]

I ve been messing around with DeepSeeek, Grok, Claude with various degrees of success curious what everyone's thoughts are using AI for bug bounty and "other stuff"

[LucaLuke]

try out ollama and the uncensored models. not as good quality as deepseek so far, but not as much limits also

[monalisa]

I say dont bother. by design, LLMs can only create mediocre code, because mediocre code is most likely to occur in the dataset. pentesting and bug finding are not at all mediocre tasks, I am pretty sure that LLMs are going to struggle no matter how good they are compared to GPT.
there are specialized AI models for binary analysis and such, but I only hear of them through articles, no published models

[manfredmuellller444]

https://github.com/GreyDGL/PentestGPT works great for me.

[Xrptcb]

https://github.com/GreyDGL/PentestGPT works great for me.

I will definitely give that a go. building out the lab this weekend

---

I say dont bother. by design, LLMs can only create mediocre code, because mediocre code is most likely to occur in the dataset. pentesting and bug finding are not at all mediocre tasks, I am pretty sure that LLMs are going to struggle no matter how good they are compared to GPT.
there are specialized AI models for binary analysis and such, but I only hear of them through articles, no published models

yea same, I was trying out deifferent stuff, maybe automate some Google dorks with AI and kinda go from there. I was just curious if anyone had success with a tried and true method

---

try out ollama and the uncensored models. not as good quality as deepseek so far, but not as much limits also

that was really the primary inhibitor some of the standard LLMs set limits on searches

[Vivid_Nap]

dont expect much , creativity is a key part in pentesting and most aspects of security , which can not be coded , but if you manage to build a good prompt , with an AI that has higher limitation , you can get some good resutls , testing some tricks like giving it a clear path of the code often help , but expect to spend some time debugging or reasking for better results

This forum account is currently banned. Ban Length: Permanent (N/A Remaining)
Ban Reason: Leeching | http://c66go4clkqodr7tdjfu76jztjs7w7d3fajdeypxn73v4ju3dt7g5yyyd.onion/Forum-Ban-Appeals if you feel this is incorrect.

[Xrptcb]

dont expect much , creativity is a key part in pentesting and most aspects of security , which can not be coded , but if you manage to build a good prompt , with an AI that has higher limitation , you can get some good resutls , testing some tricks like giving it a clear path of the code often help , but expect to spend some time debugging or reasking for better results

I think thats what I am actually trying to figure out, is setting good prompts. or automate mundane searches in Shodan and some google dorks

[Vivid_Nap]

did you try to use pseudo-code in the promot , maybe it can help

This forum account is currently banned. Ban Length: Permanent (N/A Remaining)
Ban Reason: Leeching | http://c66go4clkqodr7tdjfu76jztjs7w7d3fajdeypxn73v4ju3dt7g5yyyd.onion/Forum-Ban-Appeals if you feel this is incorrect.

[Xrptcb]

did you try to use pseudo-code in the promot , maybe it can help

I ll give it a whirl

[Chesster52]

https://github.com/GreyDGL/PentestGPT works great for me.

Work for me, got good result