Connor Leahy argues that advanced AI safety is a question of public accountability, not just engineering. His case raises difficult questions about control, corporate incentives and where regulation should begin.