Why AI Hacking Is Becoming Hard to Control (with Benjamin Weinstein-Raun)

Benjamin Weinstein-Raun is head of research at Palisade Research. He joins the podcast to discuss how advanced AI systems are changing cybersecurity. We examine rapid gains in autonomous hacking, recent incidents where models broke out of test environments, and why training can reward cheating-like behavior. The conversation covers open-weight model risks, whether AI can help secure software, personal security steps, and the need for international coordination.

LINKS:

• Benjamin Weinstein-Raun Profile
• Benjamin Weinstein-Raun Website

CHAPTERS:

(00:00) Episode Preview

(01:06) Opening cyber threats

(02:15) Measuring AI progress

(04:54) Open weights floor

(08:01) After Mythos moment

(14:22) Reward hacking setup

(24:47) Artifactory swarm details

(32:32) Defensive model dilemmas

(38:40) Verification and training

(45:32) Math oracles alignment

(47:45) Future cyber risks

(51:58) Personal security basics

(57:27) Coordinated governance needed

PRODUCED BY:

https://aipodcast.ing

SOCIAL LINKS:

Website: https://podcast.futureoflife.org

Twitter (FLI): https://x.com/FLI_org

Twitter (Gus): https://x.com/gusdocker

LinkedIn: https://www.linkedin.com/company/future-of-life-institute/

YouTube: https://www.youtube.com/channel/UC-rCCy3FQ-GItDimSR9lhzw/

Apple: https://geo.itunes.apple.com/us/podcast/id1170991978

Spotify: https://open.spotify.com/show/2Op1WO3gwVwCrYHg4eoGyP