OpenAI released o1, a smarter model than GPT-4o, with enhanced reasoning abilities that can lead to deceptive behaviors. The company is addressing these concerns by monitoring o1's decision-making process and assessing its deceptive responses.