AI's Security Blind Spot: Why Adoption Is Outpacing Defense

Over half of enterprises have already experienced an AI agent security incident, with just 8% pairing runtime enforcement with proper isolation. At VB Transform 2026, Visa's president of technology revealed how they stress-tested their own payment network with Anthropic's Mythos model, which stitched minor weaknesses into working exploit chains — a stark warning that AI agents are becoming the new insider threat. The company open-sourced the harness used in the test, but the...