William Gibson had ICE, Cyberpunk 2077 has the Blackwall. All tasked with keeping rogue AIs from breaking through. Sci-fi's answer to rogue AI seems to be to wall it off. The reality is that an AI cut off from data & purpose is useless. We can't firewall our way out of this one.

Nearly every photo id a KYC verification asks for is online & can be used as an AI prompt. Before, the word "specimen" WAS the security control. Now it's a handy variable telling LLMs where to infill. Pre-AI era controls are now on par with "I sent the password in 2 emails."

For any aspiring hackers, please stay ethical it really isn't worth it.

I have deleted the airplane mode button and replaced it with a shortcut with an airplane on. Press it and it’s photographs the user, locks the phone and turns WiFi etc on.

Goodbye zero-days, 2026 is the year of the zero-hour exploit.

Most CTFs are brute-forceable deterministic escape rooms that do not teach tactics, stealth or deception b/c there is no active IR or monitoring ergo very little transferable skills towards cyber defense. We need more defenders & builders that know secure arch, not pentesters.

Periodic reminder: Good security architecture minimizes the number of bugs and CVEs you have to care about.

The problem is that real skills are priced out of the market b/c the demand side cannot tell the diff. anymore between pros that know how to build/secure, & tool runners. Orgs are recruiting their security staff out of the PM & pentest pool so compliance & insurance run supreme

You guys, I JUST found out about recency bias, its my favorite thing ever

Sign of the times

Say what you will about Windows (OS), but the NT kernel really is an engineering marvel that still puts Linux to shame in many ways. The quickest way to describe it for a programmer, is NT was more like an object-oriented language, with a strong security model from day one, whereas Linux is very…not. A lot of the “good” features in Linux feel bolted-on (SELinux, Capabilities, Namespaces) because…well they were. I love to imagine an alternate history where NT won. IMO, Microsoft *should* have made an “Open NT” in the early 2000s; not fully GPL-style open, but one where a large org could say…swap out a memory allocator for their own. (they sorta did this with limited source access, but it was too restrictive) You could imagine say…an early Amazon forking OpenNT to create an “AmazonNT” for EC2, where they have a modified scheduler, network stack, whatever. But, the security+compatibility contract keeps a stable baseline on the Microsoft side. Controversial take, but if we enter this era where users are giving AI agents increasingly higher levels of access control; the Linux kernel is legitimately a poor fit. Think about it; answer the question “What exactly is this AI agent allowed to do?” On Standard Linux, it’s disgustingly messy with lots of overlap. Do you use UIDs? GIDs? ACLs? CGROUPs? Policies? SELinux? Filesystem modes? There’s not a singular coherent graph of capabilities you can point to. Too many ways you can escape an initially narrow scope. NT, by comparison, can go the route of explicitly typed resources, and then you could have these really strong centralized audit trails when an agent goes haywire…etc. I know I’m rambling, but the point is…if you were greenfielding an OS kernel from scratch, in 2026, with the intent of being forward-looking, it would *not* look like Linux. Frankly, it’d probably look a lot closer to NT, or even a BSD fork…

Before LLMs we had Dr. SBAITSO & when it didn't know the answer it made no issue in disappointing you, which was often. That is before hex-editing the binary & boobytrapping the answers for our friends for a laugh. We had to write our own hallucinations in the 90s.

When websites still had personality and weren’t just a bunch of div blocks with a hamburger menu

Zero-day remote code execution vulnerability in iPhone Safari. Click a link, and your crypto, passwords and everything else on your iPhone are gone. Exploited in the wild by "DarkSword" malware. "The DarkSword attack program has leaked, with its core capability being: extracting forensic-level data from iOS devices via HTTP interfaces. In actual attacks, attackers can combine social engineering or watering hole attacks to lure users into falling victim, thereby stealing data from iPhone / iPad devices and uploading it to servers controlled by the attackers." Update iPhones immediately. Apple originally patched this, but rumours suggest even the latest versions are vulnerable, “pending confirmation,” across a wider range than the original 18.4–18.7 window. From a Chinese security researcher, SlowMist CISO, @im23pds nitter.net/im23pds/status/2101265…

The combination of the #Brucon conference and #Corelan training is about the best price/quality offering you can get right now for realistic exploitation training. I am sure it will be.. *looks around* HEAPs of fun! 😶‍🌫️

On July 25, we hacked OpenAI. Two bugs let us take over ChatGPT/Codex accounts of OpenAI employees (+some unaffiliated users) and reach connected services: Outlook, Slack, GitHub, etc. We proved it with a PR in OpenAI’s internal codebase . It took us <72h. 🧵

I just read something about LLMs self exfiltrating their weights via CPU temperature covert channels. I think people should ask their LLMs how computers, physics, and information theory work.

I agree but also: (And previous incidents too) (Nobody seems to care) (Preventable incidents will continue forever)

Corporate Infosec sends a phishing test email, I click, and somehow I'm the asshole because "[I] failed the test; had this been real it would have destroyed the company network"? If me clicking can destroy the network, I'm not the one in this conversation who sucks at their job.

I’m not a lawyer, but Sam Altman and Dario Amodei may be creating a legal problem for their companies. If their AI products ever do cause serious harm, these public warnings of theirs could be presented as evidence that the risk was foreseeable and they knew about it.