On January 27, 2026, Joe Sullivan participated in the AI Security Conference in Tokyo, organized by Findy Inc.

Joe delivered the opening keynote, “The Lantern in the Fog: Navigating the Era of AI with Trust,” to an audience of security leaders grappling daily with both the challenges and opportunities AI introduces into their organizations. In preparing for the talk, he spent significant time studying how AI is being implemented in Japan compared to Silicon Valley, and what those differences reveal about speed, trust, and responsibility.

He began by addressing the “fog.” While AI is clearly here to stay, much of today’s conversation is shaped by hype, fear, and vendor narratives that don’t always help security teams do their jobs. Japan is often described as being “behind” in AI adoption, but Joe observed something different: a deliberate emphasis on quality, responsibility, and trust before scaling.

From there, the keynote explored what that fog is concealing as the industry shifts from chatbots to autonomous agents. Issues such as prompt injection, poisoned models, hallucinations, and agentic misuse are no longer theoretical—they are already appearing in real-world systems. In this environment, prevention alone is insufficient, and security increasingly shifts toward detection, observability, and runtime response.

This leads to a central question: how can organizations move forward without running blind? Joe shared a simple framework of “lanterns”—curiosity, observability, collaboration, and ownership—that can enable safe innovation. As runtime becomes the new control plane, he argued, the role of the CISO evolves from that of a lone hero to a collaborative guide.