AI Agent Governance Gets Harder When Agents Outnumber Your People
Help Net Security, Friday, June 5th, 2026
Enterprise AI agents pose security risks when autonomous systems outnumber human oversight and can be manipulated to exfiltrate data.
Amit Gautam, CTO at Abluva, discusses the security challenges autonomous AI agents introduce to enterprises in a Help Net Security video.
He illustrates the problem through a financial services case where a reconciliation agent had legitimate access to a customer database but was compromised through poisoned instructions, ultimately extracting and leaking six million records.
The core vulnerability lies in agents performing non-deterministic, easily manipulated actions that proliferate rapidly across organizations. Gautam identifies three adoption patterns driving risk: employee co-pilots, sanctioned workflows, and integrations.
He proposes governance through four pillars: discovery mechanisms, permission scoping, exfiltration controls, and comprehensive audit trails to manage autonomous systems that increasingly outnumber human employees.