The article reports on a new study by the Cloud Security Alliance highlighting a major emerging challenge in enterprise security: organizations are struggling to differentiate between actions performed by AI agents and those carried out by humans. According to the findings, about 68% of organizations cannot clearly distinguish AI activity from human activity, even though 73% expect AI agents to become essential within the next year. This signals that AI adoption is advancing faster than the systems designed to monitor and control it.
A core issue identified is the failure of traditional identity and access management (IAM) systems to keep pace with autonomous AI. As AI agents take on more responsibilities—such as automation, research, and system monitoring—existing security frameworks struggle to assign, track, and verify their identities. This creates confusion around accountability, making it difficult to determine whether a human or an AI system initiated a specific action.
The article also emphasizes the growing risk of overprivileged access. Many AI agents are granted excessive permissions or inherit access rights from human users or shared system accounts. In some cases, they operate under human identities or generic service accounts, which further blurs accountability and increases the attack surface. This lack of precise control means AI agents can unintentionally access or expose sensitive data, creating serious security vulnerabilities.
Ultimately, the study concludes that organizations must rethink their security models for the “agentic AI” era. This includes implementing clearer identity frameworks for AI agents, enforcing least-privilege access, and improving monitoring and governance. Without these changes, enterprises risk losing visibility and control over increasingly autonomous systems, making AI-driven environments harder to secure and manage safely.
