Sodium comments on Large Language Models can Strategically Deceive their Users when Put Under Pressure.