Over-Privileged AI Drives 4.5 Times Higher Incident Rates

Teleport's 2026 survey of 200+ US infrastructure security leaders finds that while AI deployments improve capabilities (faster incident investigation, better documentation, higher engineering output), they introduce significant risks tied to identity and access: 70% of AI systems have broader access than comparable humans, 35% reported at least one AI-related incident, and reliance on static credentials correlates with higher incident rates. The report recommends implementing least-privilege for AI systems, reducing static credentials, and reshaping identity teams to include platform and engineering stakeholders.