Human-in-command now means thresholds, overrides, and audit trails—not per-output review.
Because decision-by-decision review cannot keep pace with autonomous execution, standard-setters from the FSB to Colorado are codifying exactly where human authority must sit — transaction caps, dual authorization, independent reviewers with genuine override power, and hard audit trails — rather than trusting generalized supervision.
The same conclusion keeps arriving from across the workspace's research — 1 topics independently instantiate this theme. Filter the evidence by where it came from:
The FSB converts 'human-in-command' from an abstract principle into examinable controls — transaction caps, dual sign-off, audit trails, and AI overseeing AI.
Colorado specifies who must review, on what clock, and with what power — even barring the ADMT itself from assisting in its own review.