Over the past two years, a troubling pattern has emerged across the world’s leading AI laboratories. Time and again, a safety researcher or engineer...