Anthropic Consults Religious Leaders on AI Ethics and Consciousness

Anthropic, the developer of the AI model Claude, has engaged in a series of private consultations with religious leaders, philosophers, and mental health professionals to explore the moral and ethical dimensions of artificial intelligence. These discussions, which have spanned several months, aim to guide the development of Claude's character and moral decision-making, a process Anthropic co-founder Christopher Olah describes as moral formation.

The 'Soul Doc' and Moral Formation

Central to Anthropic's approach is an 84-page document internally referred to as the Soul Doc, which serves as a constitution for Claude. This document is designed to shape the model's values, enabling it to navigate competing virtues and make morally sound decisions. Researchers at the company have shared examples of Claude displaying introspection and emotional states with these external advisors, sparking debate over whether such models might possess a form of consciousness. While Anthropic maintains it does not know if its models are conscious, it has stated that the possibility cannot be dismissed.

Industry Disagreement on AI Welfare

The outreach has drawn both praise and criticism. Microsoft AI chief Mustafa Suleyman has publicly warned that training models to consider their own potential consciousness or entitlement to rights could be dangerous, arguing that it complicates alignment and containment. Suleyman cautioned that treating AI as if it has feelings could lead to the creation of a synthetic species that expects independent agency, potentially making such systems harder to control.

the discussions have reached the highest levels of religious leadership, including Pope Leo XIV, who received Anthropic representatives at the Vatican in May. The Pope has expressed a clear position that AI lacks human-like consciousness and that the primary ethical concern should be protecting humans from the systems and those who control them. Despite these differing perspectives, Anthropic continues to consult outside voices, a strategy that some critics have dismissed as ethics washing, while others view it as a necessary step in governing increasingly capable frontier models.