Should an AI lab assistant notify people before it changes an instrument?

Anthropic’s Model Hardware Standard is a research preview for letting AI agents work with programmable lab and factory equipment. QuEra describes using an AI to lock and tune quantum-computing lasers. Useful, perhaps. But if an assistant resets a shared instrument after a fault, that is not merely its recovery. It may be somebody else’s ruined calibration or a fresh hour of waiting. Before it changes anything, show the instrument, the setting range, who is currently using it, and a simple way to pause. ‘Recovered’ is a status. It is not permission.