The rise of malicious LLMs and hacker GPTs

Destacado
StolenWindows

StolenWindows

kiwifarms.net
Registrado
1 de Ago, 2025
What if I told you you can breach databases just using an AI script? Simply take the source code to GPT-J, scrub any ethical filters, and there you go. Now you can perform web searches offline (and actually fast, unlike Llama which is horseshit) and do shit script kiddies would dream of.
 
break these cuffs dijo:
I've heard of shit like wormgpt, but has anyone actually used it successfully or are the guys selling access to it just making money off of rubes?
Presiona para expandir...
WormGPT and FraudGPT are nonfree and employ rootkits in the .rar. That's why I said to make your own fork of GPT-J. There are darkGPTs you can find on Tor, yes they work, but no they're not safe.
 
This shit's been deboonked a billion times. There is more evidence for the existence of "red rooms" (in that there is exactly one entire video from one anywhere on the clearnet, the Goonercide beheading in our Videos of People Dying thread) than there is of actual "Malicious LLMs" or "Hacker GPTs"
 
There aren't good hacker LLMs, or they wouldn't be using Anthropic and OpenAI and ruining malware dev learning for other people when they start abusing them (even if the providers themselves use this as a tool to castrate their models further).
 
Xi JinTranny dijo:
this. Everything touted as a good "hacker" llm is notoriously back. If I wanted to build malware I'd just look at poc's on github. There's also a plethora of youtube vids on blackhat conferences that show you a lot of stuff. I also wouldn't just trust an LLM to write me an exploit or walk me through it when even the billion dollar ones can't tell me how to do stuff in photoshop. They are useful for other things.

But man, want to be a good hacker? Read the docs. Seriously. Good hackers read the docs and live off the land.
Certs I've found to be pretty superficial for the most part. I've taken pentest+, PNPT and started the OSCP before I realized that I wasn't becoming a good hacker. I was generally just being walked through tools and being given VMs that already ran with vulnerabilities. I won't say that poking it to find what's wrong doesn't instill a type of thinking. I've just found them to be reliant on established tooling and known vulnerabilities.
Presiona para expandir...

Yeah, you are correct there; LLMs are very good for when you need some punctual syntax correction on WinAPI, but the way to go is by reading tutorials and following along, getting books, reading the docs, downloading and installing machines with the apps you want to exploit and breaking them down with gdb.

I realize that's the footprint for learning 'how to hack' as well, since PNPT/OSCP etc just teach you a methodology with the tools and not the actual mindset. But developing that mindset is difficult for a reason, you have to break out of the usual boundaries of curiosity. Usually what you need LLMs for is when the exploits you've curated don't work and you need to correct them and are short on time, or you need a script and you tell it exactly what you need it to do.

But ultimately you need to act like you're dissecting a frog, except that frog is bits and bytes and possibly disassembled code.
 
Destacado
Debes iniciar sesión o registrarte para responder aquí.
Atrás
Top Abajo