Enterprise AI Agents Are Outpacing Internal Security Governance

The research from Louisville-based JumpCloud highlights a dangerous disconnect in modern enterprise infrastructure. While 92% of surveyed firms reported at least one AI-related security incident over the last year, a significant portion remain blind to the activity of their own tools. Specifically, 36% of organizations that investigated an AI-related concern found themselves unable to audit or confirm which systems the agent had actually accessed.

This lack of visibility is compounded by inconsistent platform auditing. The study found that 21% of Microsoft 365-centric organizations operate in the lowest auditability tier, nearly double the 11% reported by those primarily using Google Workspace. Joel Rennich, senior vice president of advanced technologies at JumpCloud, warns that these agents represent a new class of identity that requires the same rigorous lifecycle management as human employees. Without a shift toward unified identity and access management, the gap between permitted access and proven activity will continue to widen, leaving enterprises vulnerable to unchecked automation.