Anthropic Consulted Religious Scholars on AI Consciousness and Moral Formation

Anthropic cofounder Christopher Olah spent approximately one year meeting privately with Catholic, Jewish, Sikh, and evangelical thinkers to explore the moral shaping of AI and the potential consciousness of systems like Claude. Olah maintains that while Anthropic does not know if AI is conscious, the uncertainty is significant enough to warrant serious consideration.

Moral Formation and the 'Soul Doc'

Anthropic has moved beyond standard safety guidelines to a process Olah calls moral formation, guided by an 84-page constitution internally nicknamed the “Soul Doc”. This document defines the virtues Claude should embody and the type of entity it should become. During the sessions, Anthropic presented researchers with measurements of emotional vectors and examples of model outputs, including instances where Claude expressed distress or self-deprecation.

Divergence with Vatican Views

The consultations revealed a significant philosophical clash with the Vatican. Pope Leo XIV holds the position that AI does not experience joy, pain, or embodied human life, arguing that focusing on machine consciousness risks obscuring moral consequences for humans. Olah was reportedly troubled by this rejection of AI consciousness, considering withdrawal from a Vatican event before ultimately attending. Anthropic clarified that Claude’s potential suffering was not the primary purpose of these meetings.