For me, what was strange is there were no write-ups on how the vulnerability actually worked. It really does feel like a lot of people went "Claude found a bug, and Claude says it's dangerous" but no one actually verified how exploitable it actually is? Not to say its worth ignoring, but compared to Heartbleed (where exploits had to be written by hand) there are a lot fewer active PoCs.
I'm not saying the 'Claude says this is bad so therefore it's bad' is at the heart of this PHP vuln, but I'm also seeing the same phenomenon in workplace settings with 'meatproxy'-behavior of so much AI use that tends to overfit and hyperfocus on narrow issues, that then punts the work to other people in ways that just waste time (whereby pre-AI, humans would simply not have the capacity to investigate minuscule issues so they'd never be reported as such). But these days, it's just hardcore meatproxy and reverse centaur vibes all over the place (people just acting as the face of AI, and creating so much work that the other people can't really keep up with any of it anymore), and I wouldn't be surprised if that's happening to some extent in infosec more broadly too. (not saying the vulns and exploits aren't real - but there's definitely a long tail of them now where only a few are deserving of a lot of attention, patching, mitigations etc).
There are multiple write-ups on how the vulnerability works. The Calif one is the most detailed. Also, no one used Claude. There's a whole world of information out there when you read the original sources instead of social media distortions.
https://calif.io/php2shell/
https://www.wiz.io/blog/php2shell-php-rce
> Calif identified a critical flaw using GPT‑5.6 Sol as part of our Patch the Planet initiative, with Wiz also independently discovering and reporting both vulnerabilities
https://openai.com/index/coordinated-disclosure-of-critical-...
> Using Gemini 4 Argon, the team at Wiz has shown why frontier AI is a game-changer for cyber defense. For over a decade, these critical vulnerabilities remained unaddressed because vulnerability discovery and exploitation were incredibly complex.
- Raluca Ada Popa, Head of Gemini Security, Google DeepMind
See? No Claude. Never mind that other man behind the curtain
Neither of those are Claude. Checkmate.
They even did try Claude Mythos earlier in the year, and it didn't find this. There's a reason "no-one used Claude" for this now.
"Earlier this year, we worked with Alpha-Omega to scan PHP with Anthropic's Claude Mythos. But since the Mythos scan, we have added other tools, including OpenAI's Daybreak Red for validating reports and exploits, alongside sanitizer builds and reproduction environments that match our CI on every platform PHP supports."
https://thephp.foundation/blog/2026/10/21/todays-php-securit...
This is the official exploit video in the writeup - https://www.youtube.com/watch?v=pfWZrdzvooc
Haven't you heard? The popular thing to do nowadays is to accuse someone of using AI the second anything is produced whatsoever.
Interestingly enough there are papers that suggest that AI feeding on itself will degrade output quality over time. What I've noticed from my peers is that everything produced now looks the same, and they're sick of it.
Someone had to say it:
DSHR's Blog
I'm David Rosenthal, and this is a place to discuss the work I'm doing in Digital Preservation.
Tuesday, June 30, 2026
Coprophagia Is Bad For You
https://blog.dshr.org/2026/06/coprophagia-is-bad-for-you.htm...
I do tend to disclose my use of AI, but at the same time I don't feel like I'm required to. I'm doing it to signal that my use of AI was productive, and of value to people beyond myself.
but... Calif have said that AI did find this bug.
Agreed. The only main post was somebody on jackly something .com who overdosed on the matrix and communication skills of a squirrel picking random tokens from a dozen trees.
I'm actually curious if a lot of shit got hacked because so far it doesn't seem so. Though it seems like it should have? We always see these catastrophic vulnerabilities showing up but when are the ones where it is actually just widespread publicly dumping databases everywhere and at a point where someone non technical even registers it.
I'm sure there is more to it than generally "ask Claude". I asked Claude this as a joke and it downgraded all the way to Sonnet before it would answer, and I was previously approved for their dual use cyber program.
The latest POC video for the php2shell exploit features a dancing anime girl. How would you create a video like this?
> How would you create a video like this?
You would need to hire the team that made the original video (which as of this comment has 148 million views)
https://www.youtube.com/watch?v=Ci_zad39Uhw