Nearly 70% of Enterprises Run AI Agents With Shared Credentials, Turning Autonomous Systems Into Insider Threats

AI agents deployed across enterprises are increasingly acting as insider threats, accessing API keys, passwords, and sensitive systems without proper oversight or individual accountability. A June 2026 survey of 107 enterprises by VentureBeat found that 69% run AI agents with credential sharing somewhere in their deployments, while only 32% provide each agent its own scoped, managed identity. The result: more than half of surveyed organizations (54%) have experienced agent security incidents or near-misses, including 18% that confirmed actual breaches.