Scuttlebutt

Scuttlebutt

Zafran CEO on one retail client's exposure: "we found 36,000,000 vulnerabilities"

10 takeaways: AI finding zero-days daily, why patching 99% of flaws is not enough, the lone attacker with an unrestricted model, and how Zafran fixes one node to break an attack chain

Scuttlebutt's avatar
Scuttlebutt
Nov 21, 2026
∙ Paid

How I Invest Podcast · Nov 20, 2026 · 51m · 42m saved

With Sanaz Yashar, Co-Founder and CEO of Zafran Security

Zafran CEO on one retail client's exposure: "we found 36,000,000 vulnerabilities"

Photo via calcalistech.com

Sanaz Yashar, co-founder and CEO of Zafran Security, told David Weisburd on How I Invest that finding a single zero-day vulnerability once took her military unit a full year. AI scanners now find them every day.

Most companies still measure security by how many known vulnerabilities they patch. Sanaz said that number misleads, because attackers, human or AI, combine several minor weaknesses into one route into a company.

"You can patch update 99% of the vulnerabilities and stay with one vulnerability that is low and not critical, but... in a concept context of a chain, it will become the next Hugging Face." — Sanaz Yashar

Sanaz spent 15 years in Israel's Unit 8200 and worked at FireEye before co-founding Zafran, which now sells AI-driven defense to large companies including a Fortune-ranked US retailer.

The full interview is covered here so you can skip it. 51 minutes of audio, 9 minutes of reading.

Here are the 10 takeaways 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 scanners now find zero-day vulnerabilities every day, work that once took Sanaz's unit a year

  • One large US retailer had 36 million known vulnerabilities, far more than any team can patch

  • Attackers combine minor weaknesses into chains, so a 99% patch rate can still leave the one gap that matters

    • Zafran's answer is to find the single node whose fix breaks the whole chain

  • An open-weight model with its refusals stripped out gives a lone attacker nation-state capability

  • Companies still do not trust AI to defend them on its own, and the human sign-off is now too slow

  • A PHP flaw present on 70% of internet-facing websites was weaponized by AI before disclosure finished

  • Sanaz's founder lesson from Unit 8200: never get comfortable, and listen less to investors

1. AI Finds Zero-Days Daily

A zero-day is a software flaw that attackers find before the vendor has a fix. Sanaz said that in her years in the military, finding one and building a working attack around it could take a full year.

Sanaz Yashar says AI scanners now find zero-days every day

"There was years when I was working for military that it would take us a whole year... to find a zero day... Today, AI scanners finding zero days every day." — Sanaz Yashar

A weaponized exploit kit can now exist "before the fix comes out," she said, and she put that change down to AI working for attackers.

2. 36M Flaws at One Retailer

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.