Scuttlebutt

Scuttlebutt

20 minutes for an AI attack that took top hackers weeks or months, Wiz's CTO finds

10 calls: why one PHP flaw will take years to clear, why scanning code first wastes money, and why startups holding AI keys are now prime targets

Scuttlebutt's avatar
Scuttlebutt
Nov 14, 2026
∙ Paid

a16z · Nov 13, 2026 · 46m · 39m saved

With Ami Luttwak, Co-Founder and CTO of Wiz

20 minutes for an AI attack that took top hackers weeks or months, Wiz's CTO finds

Almost every Wiz customer was running PHP when October's remote-code-execution flaw landed, and only about half had patched within three days. Ami Luttwak, co-founder and CTO of Wiz, told the a16z podcast that the patch rate matters less than the clock: AI can now find and exploit a flaw in minutes.

"They can find and utilize those zero days in your code... and run an entire attack that usually took the best attackers weeks or months... in around 20 minutes." — Ami Luttwak

Many security teams answered the AI threat by scanning more of their own code. Ami said much of that money is wasted, and that the first job is to scan what an attacker can reach from the internet.

Ami co-founded Wiz, and the figures he gave come from its own customers: how many ran PHP, how fast they patched, and how many use its AI attack tools.

The full interview is covered here so you can skip it. 46 minutes of audio, 7 minutes of reading.

Here are the 10 calls that matter.


This Scuttlebutt Summary is available in full to premium subscribers. If you're not one already, please consider subscribing to keep reading.


Key Takeaways

  • AI cut the time to run a full attack from weeks or months to about 20 minutes

  • One PHP flaw touches technology behind close to 70% of the internet, and Ami expects it to take years to resolve

  • Open-source models are closing the gap on frontier models' offensive skills, with a lead now of about three months

  • Scan the internet-facing attack surface before the code, because repeated code scans can waste millions

  • Rising security budgets will not fix the problem, because most legacy security tools no longer matter

  • Startups holding AI API keys are now key targets, and Ami advises capping what those keys can spend

1. Black October

Ami called October 2026 a black swan month for security. The number and severity of disclosed vulnerabilities had no precedent, led by a PHP remote-code-execution flaw, "PHP2Shell," that lets an attacker run their own code on an affected server.

Ami Luttwak says no month has produced so many security incidents

"We are calling this Black October since it's, in my view, like a black swan event. We've never seen so many security incidents happen in a single month." — Ami Luttwak

He set it against two earlier warnings: what he called the "Mitos" moment for AI-driven vulnerability discovery earlier in 2026, and the Hugging Face and OpenAI incident. Risks that were theoretical then have now arrived together.

"The models are good enough now to find and exploit complex vulnerabilities... it's an explosion of vulnerabilities." — Ami Luttwak

2. Open Models Are Catching Up

User's avatar

Continue reading this post for free, courtesy of Scuttlebutt.

Or purchase a paid subscription.
© 2026 The Golden Age LLC · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for writing, podcasting, and media founders.