Anthropic's Pilot Sabotage Risk Report — AI Alignment Forum