Can Shadow AI Risks Be Stopped?

Enterprises are increasingly granting nonhuman identities broad, sensitive access as the AI agent landscape expands, exposing fresh security blind spots around shadow AI and shadow agents. Entro Security has expanded its platform with an AI agents discovery and observability tool to stop shadow agents, aiming to map AI agents, NHIs and secrets across organisations’ stacks and improve ownership and life-cycle management.

Recent internal Entro Security research found that over 90% of NHIs operate without proper life cycle management, a factor that makes shadow AI a prime target for attackers. With developers spinning up shadow AI agents using large language model APIs from OpenAI and Anthropic, risk is escalating as these agents often hold powerful credentials yet lack governance or revocation paths.

According to Itzik Alvas, co-founder and CEO of Entro Security, the tool monitors agents’ purpose, permissions and blast radius and provides a complete view across code, cloud, endpoints and SaaS. Forrester’s AEIGIS framework also emphasises continuous risk management and mutual authentication between agents and tools to curb these risks.