
The WhatsApp complaint vs NSO contains some fun technical exhibits. The user manual and Ghana contract reveal quite a bit on NSOs system design and thinking.

There is never enough time, thank you for yours.

The WhatsApp complaint vs NSO contains some fun technical exhibits. The user manual and Ghana contract reveal quite a bit on NSOs system design and thinking.

So #java4shell was apparently noticed by some developers two months ago without knowing the significance? github.com/adoptium/adoptium… And they had a fix

One thing to eventually discuss with a therapist or LLM: I live in perpetual frustration of people making demands on my time, and annoyance about all the things I am not making enough progress on because people keep stealing my time.

in this post the Claude-Code dev team boast that 90% of claude-code code is written and verified by agents, without a human review. and then they provide some generic tips about what are good tests etc. but my take is that the real reason they can pull it off is not their amazing auto-testing skills, but the fact that claude-code is essentially a very simple piece of software, they don't have almost any efficiency concerns, no competitors, very little potential of actually harmful bugs or data corruption due to claude code itself, and that many bugs just got swallowed and compensated for by the underlying LLM being so freaking good.
Claude writes better code when it can verify its work. Checks that used to be expensive, like tests, benchmarks, and end-to-end runs, are now cheap. Here’s how the Claude Code team thinks about verification: claude.dev/blog/how-the-clau…

Follow-up question: What was inside the scope of what they were proving and what was outside? Side channel is the classic example but also multi-threading is incredibly hard to model properly inside formal verification.
One thing that I am curious about with recent formal verification of software results - how many security relevant counterexamples (ie vulnerabilities) did they find while producing the proof?

Last time this happened we moved from a very distributed internet to a more secure centralized applications

Colleagues doing IR are overwhelmed as never before and it's not because of a crazy exploit or supply chain breach. At this moment the common attack tool overwhelms the common defenses.
**We’ve been moving from the vulnapocalypse to the breachapocalypse.** Artex is the tool of the hour, so we ran it through our internal evals to understand what defenders are actually up against. Full breakdown by @ace__pace ⬇️ tenzai.com/blog/artex-the-ha…

It's a good tool. I get why it's used by criminals and low end espionage. But it's nothing special. You can vibe code this yourself

I wouldn't run Artex in a sensitive environment. The agent has access to the LLM key and admin panel, without any sandboxing. Any indirect prompt injection leads to system takeover and data exfiltration.

Artex is amazingly flexible as an agent. The customization is great when it's your setup. It's a lot less fun when you need all of those choices to work consistently. Creating tools mid run is neat, trying to debug this is not.

So how does compare? It's on the top part of my leaderboard of publicly available harnesses for web PT. Finds vulnerabilities but also finds a lot of noise and crap. This is great for hacking and great for PT but less appropiate for AppSec work

Why Artex and not Strix last year? Open weight models weren't good enough and the value wasn't there for criminals yet

Artex is getting press because it's the right tool at the right time for cybercriminals. Cheap, usable UI, capable enough. You don't need the best agent if the one you have makes it easy to get started.

So Artex is the vibe coded agent of the month, right as we move from the vulnapocalypse to the breachapocalypse. I ran it through our internal evals to see how it compares with other public harnesses. See thread

Tommorow @BsidesTLV I'll talk about the journey of BragJack, where I was able to hack 5 of the most popular browsers in the world with the same technique - should be fun!

I think small conferences are the way to go if you actually want to hear something novel and actually talk with speakers showing cool stuff Doing this specifically for AI Offsec is new!
In London for Black Hat EU? Request a seat to hear the top builders, practitioners, and researchers in the offensive AI community trade hard-won lessons. Tuesday, December 8, in the underground Churchill War Rooms. Invite-only. Request yours at misalignedcon.org