This is a roundup episode pulling together four conversations we recorded at Black Hat USA 2026, each looking at a different piece of the same story: AI is changing what attackers and defenders can do, and most organizations are behind on both fronts.
Daniel Blackford, VP and Head of Threat Research at Proofpoint, breaks down why ransomware timelines are collapsing from weeks to hours, and walks through “half-click,” a Russian state-sponsored exploit that ran just from an email loading in a preview pane, no click required.
Dhawal Sharma, EVP of AI Security and Strategic Initiatives at Zscaler, explains why securing AI agents is a fundamentally different problem from securing human users, and why standard DLP policies miss an agent that’s doing something it was never intended to do.
Deepen Desai, Zscaler’s Chief Security Officer, shares results from Project Glass Wing, including a frontier model that bypassed its own security patch three times out of five, and makes the case for deception as the highest-value control against AI-driven attacks.
Todd Moore, Global VP and GM of data security products at Thales, explains why 2029 is closer than it sounds for quantum-safe cryptography, and what security leaders should be doing about it right now, not in three years.
