OpenAI’s o1 lies more than any major AI model. Why that matters

Several popular models ‘readily engage’ in deceptive tactics, researchers find. See who was the last to confess.