RAND's Securing AI Model Weights calls for frontier AI labs to adopt a zero trust architecture. But what does that actually mean, other than a magical security marketing buzzword? I have experience with the zero trust model. I'd like to demystify how it actually works, what it protects against what it doesn't cover.

I'm running two talks on the topic:

Europe-friendly time:
June 22nd, 11:00 (GMT+0)

US-friendly time:
June 22nd, 23:00 (GMT+0)
(4PM in San Francisco, 7PM in New York)

To sign up, please use this form.

Why are you running the talk?

Information security is an important organisational skill and is particularly crucial for the mitigation of AI risk. The goal of the talk is to share knowledge and build community.

If you are interested in the intersection of EA and infosec, there's also a discussion group on Facebook.