AI Investigating AI: The Architecture Behind Securing Synthetic Insiders
Security Boulevard, Saturday, October 10th, 2026
Above Security's CPTO explains how to use AI to investigate AI agents as insider risks without making it the sole authority over evidence.
In an Above Security post syndicated to Security Boulevard, the company's CPTO responds to Satya Nadella's essay arguing that capable AI models should be treated as insider risks, subject to the observation, access control, auditing and containment applied to privileged humans.
Nadella's line, 'separate the supply of intelligence from the authority over it,' frames the design, resolving the apparent paradox of one model judging another by letting AI reason about behavior without being the sole authority over the evidence.
That starts with independently observing each AI agent rather than just feeding logs to an LLM.