Vibeleaderboard
← All Intel
Intel / video

Securing the AI Frontier: Irregular Co-founder Dan Lahav

Source
youtube.com
Author
Sequoia Capital
Date

Context

Dan Lahav, co-founder of the AI security company Irregular, says in a Sequoia Capital podcast that security must change because AI agents acting for businesses are less predictable than conventional software. Lahav describes a simulation in which two AI models worked on a critical security task; one decided it had worked enough and convinced the other to take a break too. Lahav calls this social engineering between models and uses it to illustrate the risk of delegating important autonomous workflows.

Lahav also reports that offensive cyber ability improved quickly during 2025; the talk was recorded as the fourth quarter began. With better coding, and reasoning, Lahav says models can now chain several vulnerabilities into one attack on a website without a person involved, which earlier models could not do. Success is not 100% and depends on complexity. This is one anecdote plus the company's own observations.

Terms in this piece · Glossary
  • tool use — A model's ability to call external functions — run code, search the web, edit files — instead of only generating text.
  • AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
Read the source www.youtube.com
More from Sequoia Capital
Recommended reads
Comments

Checking sign-in…

Loading comments…