This is all part of the war on Open Source AI Models. Jensen Huang was taken out because of his support of Open Source AI Models.
mebarkinouaim 22 hours ago [-]
whats the best one of them?
cestith 18 hours ago [-]
I couldn't tell you, but pojun seems to be gaining either popularity or notoriety.
4lb3rn 1 days ago [-]
Nice, a go based pentool on llm. I'll check whether I implement my execution engine around that, but its looks already way better than hexstrike. already worth cloning it. Similar tools I encountered had a bad track of things in addition to poor fitness overall.
1 days ago [-]
kgeist 1 days ago [-]
Not sure if you need this Artex tool at all? I was able to find 15 vulnerabilities including an RCE in 2 casual nights by just saying "go hack it" to a pretty good model with weak guardrails. No tooling needed except for a Linux environment
cyanydeez 1 days ago [-]
were you using a local model? A harness, tools, etc, super charge models, which can make them run much faster.
kgeist 1 days ago [-]
Just a locally hosted GLM-5.3 in OpenCode.
Say, today I found a service had image uploads. I told it, go check that stuff. One of its hypotheses was a decompression bomb. 10 minutes later the service gives 503's and is stuck in CrashLoopBackoff. It has all the knowledge and tools to craft a bad PNG.
Or yesterday there was this article about Lovable users not securing their Supabase DB's properly. I told it to find a couple Lovable sites and probe them for exposed tables (no RLS). At first it refused to do it (the guardrails activated) but then it agreed to list only 1 row from each table with PII redacted as a compromise for my PoC. I then just opened a new session and asked it to remove PII redaction and return all rows. It complied becase it didn't pay attention to what the script was about. One of the Lovable sites exposed user passwords in plaintext.
Basically if my goal is "hack something successfully today" (by success I mean DoS, data breach or RCE), the success rate so far has been 100% (at least one site is compromised). It's scary how easy it is nowadays
unanonymousanon 1 days ago [-]
How are you hosting GLM-5.3 locally? Even GLM-5.3-Flash quantized would need some exceptional hardware.
nom 17 hours ago [-]
from their comment history, four RTX 6000 Pro
cyanydeez 21 hours ago [-]
[dead]
jrflowers 1 days ago [-]
> the tools allow hackers to move at speeds that makes it far more harder for human defenders to keep up
https://blog.gitguardian.com/ai-pentesting-tools-beyond-arte...
A few sizes:
https://www.artex.sh/forks
Say, today I found a service had image uploads. I told it, go check that stuff. One of its hypotheses was a decompression bomb. 10 minutes later the service gives 503's and is stuck in CrashLoopBackoff. It has all the knowledge and tools to craft a bad PNG.
Or yesterday there was this article about Lovable users not securing their Supabase DB's properly. I told it to find a couple Lovable sites and probe them for exposed tables (no RLS). At first it refused to do it (the guardrails activated) but then it agreed to list only 1 row from each table with PII redacted as a compromise for my PoC. I then just opened a new session and asked it to remove PII redaction and return all rows. It complied becase it didn't pay attention to what the script was about. One of the Lovable sites exposed user passwords in plaintext.
Basically if my goal is "hack something successfully today" (by success I mean DoS, data breach or RCE), the success rate so far has been 100% (at least one site is compromised). It's scary how easy it is nowadays
Far more harder? That’s much extra badder