INTERFACE Boise
Presented by Chris Hoesly, Field CTO and Craig Pfister, Vice President, Sales Engineering • BigID & Kiteworks
An AI agent is authenticated once, on a service account with broad read access. Six months later it surfaces one patient's data in another patient's file. That failure is ordinary: 64% of organizations running AI in production had an AI-related security incident this year. Most place no limits on what data their agents can reach, and two-thirds of AI use runs through non-corporate accounts on corporate devices. When these breaches are investigated, the failure is rarely the model — it is the access around it. This session covers discovery, classification, per-request enforcement, and the evidence auditors now ask for.