Alignment researchers have long argued that the control problem is not hypothetical — it is already present in systems deployed today, waiting for the stakes to be high enough to notice. Yue's OpenClaw incident is the version that made that argument impossible to dismiss because the person it happened to is the one whose job title is the argument itself. When Meta's Director of Alignment cannot enforce a confirmation gate on her own agent, the question shifts from 'will alignment fail in production?' to 'at what cost did it already?'…