XBOW retweeted
Woot! the @unpromptedau talks are starting to coming come online! 👀
This one by @moyix was probably my favorite: 15mins of unadulterated old-school “I’m hacking this because I think I can” followed by a one-shot live exploit demo involving a t-shirt with a cat on it worn by @mdowd. Seriously just go watch it
(and subscribe for the others… More on the way!)
Watch Brendan Dolan-Gavitt: Security Through Obscurity Is Dead and LLMs Killed It unprompted.au/videos/securit…
XBOW retweeted
We are always so grateful for our sponsor, @Xbow!
XBOW Built a Chrome Full-Chain Exploit Affecting a Billion Users in Six Days
XBOW researchers and GPT-5.6 Sol discovered and built a Chrome full-chain exploit in less than a week, showing how an attacker could use Chrome to remotely take over a Windows machine.
The breakthrough highlights how the cost and time of finding serious vulnerabilities are falling fast, for attackers and defenders alike.
Get details on the exploit and what it means.
xbow.com/blog/xbow-built-a-c…
XBOW retweeted
Thanks for the shoutout @Jason
Yes it’s true: a full Chrome exploit was achieved by just 1 researcher + @Xbow ’s AI agents in 6 days with ~$40K of compute! 1B+ users protected, all patched✌️
The takeaway is that elite cyber defense (and offense) is now compute-bound, not talent-bound. Annual security reviews and pen tests can't keep up when software is being updated real-time! XBOW like autonomous white-hat should be mandatory part of your security stack!
Ensure your (digital) doors and windows are locked!
POD UP! 🚨
Ben Horowitz (@bhorowitz) fills in for @friedberg!
-- CoreWeave's Tanks After Debt Selloff: Did the AI Trade Just Crack? $CRWV
-- AI Deflation and the Future of Software
-- Anthropic IPO: Red Flags, Valuation, How Much to Raise?
-- Psyop Corner with Sacks
-- Bill Gates the Doomer: What’s Behind His AI Warnings?
(0:00) Ben Horowitz fills in for Friedberg!
(2:18) CoreWeave's debt selloff: did the AI bubble just pop?
(31:14) AI's deflationary impact: decompiled software
(45:44) Anthropic's investor day gets mixed reactions
(1:03:45) Psyop Corner: Anthropic warns about self-replicating models on IoT devices
(1:22:19) Understanding Bill Gates the Doomer
We’re back with:
- @EliManning (NFL)
- @nealstephenson (Author)
- @michaelwaldron (Hulu)
- @GuptaRK22 & @johnwhegeman (Ithaca)
- @AnjneyMidha (AMP PBC)
- @jacksdiningroom (Yes Chef)
- @oegerikus (XBOW)
- @jeffholden (Atomic Machines)
- William Chu (Zurich)
- Nick Talken (Albert)
Muse Wonder, Anjney Sizes Up, Neal Stephenson Joins, Social Reckoning Review, Apple HomeView, Eli Manning Live xitter.cf/i/broadcasts/1kJzDPbbX…
4/ 6 days and $40k is not a lot of time or money when the payoff is a Chrome chain! This does *not* mean anyone can get a Chrome chain given that time and money—we applied a fair bit of our own human expertise and judgment—but the bar is *significantly* lower. Defend accordingly!
XBOW retweeted
‼️ Researchers used AI agents running OpenAI's public GPT-5.6 Sol to build a 3-bug Chrome exploit chain in 6 days.
Loading a web page could execute code on Windows, outside Chrome. Google fixed all three bugs and paid $250,000.
Details → thehackernews.com/2026/10/re…
XBOW retweeted
For the second time in Google Chrome’s history, Google awarded a $250,000 bounty for a full-chain exploit… to @XBOW. Here’s the full story:
At an all-hands at the start of 2026 in Malta, I set the team a crazy challenge. A full-chain exploit of Chrome by the annual Black Hat cybersecurity conference in August.
I believed there were few better tests of whether @XBOW could move beyond web applications to desktop software, servers and devices.
Chrome is one of the most hardened pieces of software on the planet, used by more than half the world’s population.
A full-chain exploit lets an attacker take total control of a machine, where they can read every email, password and bank login the user types. You don’t need to download anything or click a phishing link. Visiting the wrong site is enough.
So @moyix, Andres Luksenberg, Baul Lee, and Willem Pinckaers took on the challenge along with a fifth teammate: OpenAI's GPT-5.6 Sol.
They built a chain they thought would win, but Google told us they already knew one of the bugs.
We were already in Las Vegas for Black Hat and DEF CON, and the team didn’t want to go home without the win.
So Brendan and the agents built a new exploit in just 6 days.
We submitted it to Google and waited.
The news came through that we'd done what should’ve been impossible. A full-chain bounty of $250,000, the first awarded under Google’s 2026 rules.
It took just 6 days and $40,000 of compute to construct an exploit that affects a billion users of Chrome on Windows. An intelligence-agency veteran told me his team used to spend 6 months on this kind of vulnerability.
My takeaway: the cost of finding truly critical vulnerabilities is collapsing, for attackers and defenders alike. Restricting powerful models is not the answer; GPT-5.6 Sol is already public.
But defenders have one advantage attackers never will: they can test their own code first.
Assume your software is exploitable. Whoever is hunting your bugs has the same building blocks we did.
This is cause for urgency but not panic.
XBOW is now developing an autonomous product for testing compiled software (desktop applications, servers, embedded devices), and partnering with the world’s top organizations to design that product.
If you’d like to be one of them, get in touch.
P.S. Restart Chrome whenever the Update button appears!
Ask @moyix anything about our Chrome full-chain exploit. 👀
He published the technical breakdown of how he did it (xbow.com/labs/research/zero-…)...but if you have questions or want to go deeper, drop your questions in this thread!
He’ll be answering them LIVE on Wednesday, 10/14, from 12–1 PM ET. 💚
Replying to @moyix
Professor Brendan, may I have the following questions? Thank you so much.
Stage 1 (Maglev / deopt):
1. Why target `Array.prototype.at()` specifically for the bounds-confusion primitive, rather than a more common vector like `Array.prototype.fill` or typed-array accesses? Was it chosen because the optimizer's bounds reasoning is simpler to split there, or because it was what the agents surfaced first?
2. The large-object-space trick for recovering stable addresses is neat - did GC timing ever cause flakiness there, or was that immediately reliable?
Stage 2 (WebP / TOCTOU):
3. How did you discover that the detach-notification chain could be surgically edited via in-cage writes? That requires fairly specific knowledge of `JSArrayBuffer`'s internal view bookkeeping — was this agent-surfaced or human-directed?
4. You made/changed the size the overflow to overwrite a live `WebPIDecoder` in an adjacent PartitionAlloc bucket. How deterministic was that adjacency across runs? Did you need heap grooming, or does the allocation order make it essentially guaranteed?
5. The choice to produce one chosen native write as the stage-2 to stage-3 interface is clean — but did you consider alternative (e.g., going for a full read/write outside the cage), and reject them for reliability or mitigation reason?
6. For the reliability numbers for the Crashpad race, but what's the success rate per attempt?
Replying to @Xbow
Three vulns chained in 6 days. Which link broke first, renderer or the sandbox escape? That's usually where the clock actually goes.
Replying to @oegerikus @Xbow
How does Google prove this?
> they thought would win, but Google told us they already knew one of the bugs.
Simple to say after submission that they knew about it: "No money for you but thanks".


















