Enterprise AI agents are evolving from chat assistants into autonomous actors capable of launching other agents, modifying systems, and executing financial transactions. As this shift accelerates, security leaders are asking a critical question: could AI agents become the ultimate insider threat?
The risk is structural. AI agents increasingly operate with privileged access inside corporate networks. According to CyberArk’s 2025 Identity Security Landscape survey, machine identities outnumber human identities by 82 to 1. At the same time, Gartner projects that more than 40% of enterprise applications will incorporate task-specific AI agents by 2026. This rapid expansion mirrors the “virtual machine sprawl” era, when organizations lost visibility over proliferating infrastructure.
The threat surface is broad. OWASP’s 2025 report on autonomous AI systems identifies key risks including prompt injection, insecure output handling, training data poisoning, excessive agency, and sensitive data disclosure. Excessive autonomy increases breach of blast radius. If compromised, an agent with elevated permissions can move laterally, initiate transactions, alter configurations, or exfiltrate sensitive information at machine speed.
Historical insider risk data adds context. Verizon’s Data Breach Investigations Report has consistently shown that internal actors contribute significantly to breaches. Now, AI agents are being granted employee-level access, but often without equivalent governance controls. In 2025, 68% of organizations lacked identity security controls for AI technologies despite widespread employee usage.
Security experts recommend treating AI agents as first-class identities. Core protections include:
- Enforcing least privilege and least agency
- Issuing short-lived, task-scoped access tokens
- Requiring step-up authentication for sensitive actions
- Separating conversational interfaces from approval workflows
- Enabling centralized monitoring and rapid revocation
The broader conclusion is clear. AI agents can enhance productivity, but without disciplined identity management, containment architecture, and governance, they can also multiply insider risk. Enterprise security strategies must evolve before agent autonomy outpaces oversight.
ソース：
https://www.zdnet.com/article/enterprise-ai-agents-insider-threat/
