Autonomous security agents need governance built into the architecture across reasoning separation, continuous evaluation, progressive trust, and agent-specific telemetry at scale.