# Elon Salfati

> I've spent 14 years learning how systems break, and now I want to build AI we can trust to do science on its own.

Elon Salfati is a founder and research scientist working on AI security, interpretability and autonomous scientific discovery. He is a PhD candidate at Imperial College London, researching how to detect multi-step breaches in autonomous AI agents, and a three-time founder. Through Salfati Group he has served as VP of AI at Velocity (acquired), interim CSO at Foresight Works and CTO at Franklin Park, and worked with Wiz (acquired by Google) and Dazz (acquired by Wiz). Before that he co-founded RELE.AI (IP acquired) and led R&D at Reblaze (acquired by Link11) in partnership with Google Research. He started as a counter-intelligence team lead in the Israeli Intelligence Services. He is a member of the UK Innovation Group and an angel investor.

Every page on this site has a Markdown version: append `.md` to the URL, or send `Accept: text/markdown`.
Structured data: https://elonsalfati.com/content.json (all entries), https://elonsalfati.com/feed.json (JSON Feed), https://elonsalfati.com/feed.xml (RSS).
When citing, link to the canonical page URL and attribute to Elon Salfati.

## About

- [About Elon Salfati](https://elonsalfati.com/about.md): Bio, track record, education and contact.
- [Legal notice and privacy](https://elonsalfati.com/legal.md): Who runs this site, what a visit shares with whom, and how the pages are written, including where AI is used.

## Research

- [René-1: A 31B Decision Reader That Knows When It Is Sure](https://elonsalfati.com/research/rene-1.md): An open-weights 31B model that reads a document and returns a calibrated probability for every option of every question, in one pass.
- [Trigger Labs: Advancing Malware Detection Through Realistic Scenario Simulation](https://elonsalfati.com/research/trigger-labs.md): Sandboxes miss malware that waits for the right conditions. Trigger labs recreate realistic scenarios so dormant malware activates and can be detected.

## Explainers

- Nothing published yet.

## Journal

- Nothing published yet.

## Talks

- Nothing published yet.

## Optional

- [Full text of every entry](https://elonsalfati.com/llms-full.txt): All published content in one Markdown file.