Handling adversarial errors - History — AI Alignment Forum