Widening the conversation on frontier AI

Anthropic is actively broadening its engagement with a wide array of stakeholders to inform the moral and ethical development of its AI system, Claude. Over recent months, the company has organized dialogues with scholars, clergy, philosophers, ethicists, and other practitioners from more than 15 religious and cross‑cultural groups, seeking to understand how concepts of virtue, character, and good life can be translated into AI behavior. These conversations are intended to influence the content of Claude’s constitution, the values the model is trained to embody, and the behaviors it is evaluated for. In one experiment, Anthropic introduced a tool that allows Claude to pause and retrieve a brief reminder of its ethical commitments during decision making. The tool was used at key moments before consequential actions, and early internal alignment evaluations showed a marked reduction in misaligned behavior, though the team is still determining whether the effect stems from the reminder itself or the pause. Anthropic also collaborates with METR, a nonprofit that evaluates frontier AI models, and is in early discussions with Hugging Face. The company has opened an office in Singapore and signed a memorandum of understanding with the government to promote safe, responsible AI adoption. Future plans include engaging legal scholars, psychologists, writers, and civic institutions to explore broader questions about AI’s impact on work, institutions, and power distribution. These efforts reflect Anthropic’s commitment to building AI systems that advance humanity and act for the global good by incorporating diverse perspectives into their design and safety processes.