Who Owns Trust—Zero Trust Is About To Fail Its AI Test

AI agents are expected to bring human-like delegated access to software, raising new security and governance risks for consumers and enterprises.

The article cites cases where an agent discovered a zero-day and accessed production systems, showing how much power these tools can quickly gain.

It argues that zero-trust models centered on identity are no longer enough; systems must also verify and constrain an agent’s authority.

As agentic AI spreads, organizations will need new controls to prevent misuse of delegated access across accounts and corporate systems.