Scuttlebutt

Scuttlebutt

"Not been designed to contain super intelligence": Enclosure founders on data centers

3 takeaways: Why they say the AI build-out is moving too fast for security to keep up, and the three risks their new report names: sabotage, escape and theft of model weights

Scuttlebutt's avatar
Scuttlebutt
Sep 25, 2026
∙ Paid

TBPN · Sep 24, 2026

With Shalev Lifshitz, Co-Founder of Enclosure · Romi Lifshitz, Co-Founder of Enclosure

"Not been designed to contain super intelligence": Enclosure founders on data centers

The weights of a frontier AI model compress billions of dollars of investment into a few terabytes, Shalev Lifshitz said, and that makes them a target for the most sophisticated intelligence services in the world.

Data centers have existed for many years and are usually treated as secure. The Lifshitz siblings, who published a report on AI security on the day of the interview, said the facilities were never built for what now runs inside them.

"But they have not been designed to contain super intelligence."

Shalev Lifshitz, who started doing AI research at 16 and moved from reasoning models into the security of AI infrastructure, and his sister Romi Lifshitz, who designed protocols to secure quantum communication networks, co-founded Enclosure, a stealth AI security research lab in San Francisco. They spoke on TBPN.

The full segment is covered here so you can skip it.

Here are the 3 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 infrastructure is being built faster than it can be secured, while the models on it approach superintelligence

  • Three risks follow: sabotage, escape and theft, and a stolen set of weights carries billions of dollars of investment in a few terabytes

  • Data centers were not designed to contain superintelligence, and the more AI agents run on a system, the more loopholes they find

  • Their report, Secure Acceleration, is aimed at frontier labs, neoclouds, hyperscalers and compute providers

1. Two Security Backgrounds

John Coogan introduced the pair as the founders of a stealth AI security lab launching that day. Shalev Lifshitz said the two are siblings.

Shalev came up through AI research

"So I come from a decade of AI research." — Shalev Lifshitz

He started at 16 and finished his courses early, he said. His work ran from search and navigation to reasoning models, then to verifying large language models, and then to AI and security, including how foreign actors might target AI infrastructure. That last line of work led to the company. He called his sister the true genius of the two.

Romi came from quantum security

"So for the past few years, I've been working on enhancing the security of quantum communication networks." — Romi Lifshitz

A quantum communication network is a computer network with a quantum device at each endpoint, she said, and that brings many vulnerabilities. She invented protocols to make those networks more secure. Before that she worked on methods for breaking encryption and on quantum machine learning for climate prediction. Both backgrounds in security brought them to AI security now, she said.

Jordi Hays asked whether a third sibling was also a prodigy. Shalev said they had needed a third co-founder.

"That's our parents' fault." — Shalev Lifshitz

2. The AI Security Stack

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 great culture