Autonomous offense had an absolutely breakout year (pun intended). AI can hack, that's pretty well settled in 2026. This talk is about the law underneath it. Machine-speed offense doesn't only depend on intelligence, it also depends on an oracle: a cheap signal that an attempt worked. Against a human attacker you patch. Against a machine attacker, you lie. This talk explains why, and why I think that one property governs the whole board: where autonomous offense is unstoppable, where it can be turned with deception, and where nobody has an oracle at all.
Driven by a passion for pulling complex systems apart to know how they work, James is a deeply technical security leader who bridges the gap between offensive realism and enterprise safety. Currently serving as the Global Security Operations Manager at Woven by Toyota, he directs the company's follow-the-sun Red, Blue, Threat Intelligence, and Incident Response operations across three continents. Prior to stepping into global operations leadership, James built Woven by Toyota's internal Red Team from the ground up in Japan. Holding an offensive pedigree that includes a handful of OSCE credentials, his technical focus spans browser exploitation, low-level mechanics, including engineering bespoke ransomware simulations to rigorously test enterprise resilience and processes. James is highly active in the global security community as a speaker, mentor, and CTF architect, dedicated to sharing pragmatic methodologies for securing critical corporate and product environments.
James Redmond