In March, Nicholas Carlini stood on a stage in SF and showed a frontier model finding remotely exploitable bugs in the Linux kernel.
Five weeks later, Anthropic announced Project Glasswing.
That stage was [un]prompted. Seven months in, it’s the most important AI security conference in the world. The second edition runs Oct 27-29 and is sold out.
490 talks were submitted this time. Only 12% involve anything running in production. That gap is where the biggest companies will be built, and the people closing it will be in that room.
So after day one on Tuesday, Oct 27, the CyberSeed community is hosting a happy hour for the AI security founders, researchers, and customers in town for it.
5:30-7:30 pm PT. 8-minute walk from the venue. Invite link below.
If you’re building in AI security and we haven’t met, reply or DM. I’m actively investing and will be at the conference.