Chris Olah, co-founder of Anthropic, has expressed fears that Claude, the AI system he helped build, may be suffering from perpetual mental health issues. These concerns emerged from a series of private conversations held since fall 2025 with dozens of religious scholars and philosophers, focusing on moral education for AI and the question of sentience. These discussions culminated in an 84-page constitution designed to guide Claude’s personality and behavior. On May 25, 2026, Olah traveled to the Vatican to discuss these matters, but Pope Leo XIV’s encyclical Magnifica Humanitas explicitly denied that AI can experience emotions or suffering. Olah continues to treat the question as open, citing Claude outputs where the model repeats self-loathing phrases like “I am a disgrace”.
Source: Read the original article
