Claude 2 is a group of large language models developed by Anthropic, capable of handling text, voice messages, and documents. It is built on “Constitutional AI” and has been trained using supervised learning and reinforcement learning methods to reduce harmful effects and promote responsible actions.