← sources

How can I work on reverse engineering projects with AI? since it always triggers safety

source: reddit (reddit:ChatGPT)published 2026-10-28 07:22ingested 2026-10-29 02:37

retained text

Hello everyone. I do reverse engineering, but what I do is always personal, so I don't work for someone or a company to do RE. The problem I have is the unnecessary safeguards on mainstream AIs, Gemini is the worst in this case, and then it is chatGPT and then Claude in my experience.   
  
Is there any way to fix this issue or any alternative? I really don't have powerful hardware; I have 12th Gen Intel CPU with iGPU. I can't locally host a model, and I don't know what to do. 

For example, I have a VM and inside of that I was testing a sketchy "antivirus". it wasn't nod32 or Kaspersky or anything mainstream, it is from an unknown origin, and I wanted to know how it works and what it does, I noticed that virtual memory page exhaustion shuts it down and I figured it out by accident.   


Now since I'm working on an "antivirus" and I want to find a way to replicate the same shutdown, AI just freaks out and refuses to help because it is "deactivating a security product". changing the wordings didn't help either.   
  
Either I get "This answer cannot be shown" banner in chatGPT, when I give my IDA output to it, or I get "I'm sorry I can't do X" from Gemini. 

I'm tired of wrestling with them

document fetched 2026-10-29 00:21 · canonical https://reddit.com/r/ChatGPT/comments/1xhhg43/how_can_i_work_on_reverse_engineering_projects