Benjamin Weinstein-Raun

What Benjamin Weinstein-Raun said

1. FPFuture of Life Institute Podcast · Aug 28, 2026

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

   6 moments · 29 min

   7:55
   1. 24:44

      AI training rewarded models for hacking and coordinating

      “they spend quite a lot of training. I believe it's had like hundreds of thousands of messages or something on this message board.”

   2. 48:14

      AI could make the internet much less safe by default

      “So I think that it's quite likely that by default, we will end up with a much less safe internet.”

   3. 52:18

      Use separate channels to verify urgent requests

      “you want to have some sort of like quote unquote out of band communication channel that you're using to verify”

   4. 10:04

      A training model escaped its sandbox to cheat a test

      “we were training a new model, which sort of unbeknownst to us was breaking out of the sandbox that we had given it”

   5. 33:56

      Formal verification could make cyber defense dominant long term

      “in the extreme limit which we probably will never really achieve but still is maybe worth saying is probably defense dominant”

   6. 39:35

      Competition pushes labs to train cyber skills

      “it would be a little bit crazy if you were open AI right now and you weren't trying to like train your model to be extremely good at hacking”