The talk argues that generative AI should be treated as a programming language that compiles differently on each run, requiring strong controls before its outputs can be trusted. Its central principle
The talk presents a controlled experiment in which three AI-agent swarms operate inside a synthetic “cage”: a red team attacks, a blue team defends, and a black team builds the vulnerable infrastructu
The article recounts the author’s September 2026 trip to Houston to run the AI Village at CYBR.SEC.CON, a cybersecurity conference that also hosted CYBR.HAK.CON and an AI-focused track called AI.SEC.C