The Cybervault podcast brings together senior voices in security to address the most consequential questions in the field, and the question Stuart McClure takes on here is both timely and genuinely complex: after years of AI being marketed as the future of cybersecurity, is it actually delivering?
Stuart's answer draws on an unusual vantage point. He was building AI-based security at Cylance a decade before the current wave of generative AI interest in the field, and he has watched the cycle of AI hype and disillusionment play out multiple times across different domains. His assessment is characteristically direct: some AI applications in security are genuinely delivering, some are delivering less than their marketing suggests, and distinguishing between them requires asking specific empirical questions rather than accepting categorical claims in either direction.
The episode covers where AI is producing real defensive value — anomaly detection at scale, vulnerability analysis in application security, pattern recognition across datasets that no human team could practically review — and where the hype still outpaces the evidence. Stuart is particularly pointed about the gap between what vendors claim about generative AI in security and what the operational evidence actually shows, which is consistent with the "AI Exposed" essay series he published through the same period.
The Cybervault conversation is a valuable resource for security leaders trying to cut through the noise and make sound decisions about where AI investment is actually warranted versus where it is being driven by marketing pressure and board-level anxiety about being left behind.