Replying to @HackingLZ
Agree. Than add working for governmenttal agencies.
I always think that most in private sector is fun and games. Somecases just a big CTF :) nothing compared to actually reality with rwal consequnces.
38
Replying to @domchell
I totally agree. Following AI with a wofe that has terminal cancer its frustrating that a lot is about cybersecurity, gaming and such.
Put the AI to use where it actually could do diffrence
1
68
Replying to @pdiscoveryio
This is kind of something I will talk about on Bsides Göteborg.
I built backdoored model on local hw and switched the running model trough a credential theft in a chained attack.
The backdoor is only triggered trough camera and visuals.
197
Got same more frequent but it solves with just refering" i got TAC" it tjen acknowledge it and continues
344
Replying to @tryhackme
This was soo the ideas during the 90/00 internet bubble. Cool but not usefull in rl. Lets se if this time around ppl like it
33
Maybe i am missing something. But all this talk about sovereig AI and most ppl are still running everything from hosting, defense/security, OS and so on by Non European companies. Why does it differ?
9
Running a 27B model on 12–24 GB VRAM sounds impressive, but heavy quant has a cost. If it hallucinates, writes bad code, struggles with reasoning, or performs below published benchmarks, it may not be the model. You simply be squeezing it too hard.
Model size ≠ model capability
1
16
Starting to put more stuff public at github.
Some alpha/beta code. From c2 too AI testing framework.
70
Replying to @TinySecEx
Lol. Agree. before you start just ask it to set up the infra it thinks its needs it will generate a installation script. Then it produce poc, triage it goes fast since you just ask it to write a proper report with reproduce steps
658
Using Codex more and more in research, now bridge with kimi cli, wsl, docker.
Works really well for what i am doing.
The need to have a specific commercial framework seems like its not needed.
23
I used to think red teaming was the shit.
Then I started doing more security research and finding new vulnerabilities.
Turns out, this is even more fun. :)
8
Replying to @HackingLZ
I agree with Thacker. I dumped Antropic over openai because of the blockers, which fires randomly. If Antropic doesnt change i wonder how there IPO goes.
335
