Religious scholars met with Anthropic

Co-founder Chris Olah began convening the scholars in fall 2025, mostly in San Francisco. Reports put attendance at roughly 20 people. They came from Christian, Jewish, Hindu, Latter-day Saint, Sikh and other traditions. One April 2026 dinner reportedly drew about 15 Christian leaders. Participants reviewed actual Claude outputs, including slides of the model repeating self-critical phrases. They also heard Olah describe "emotional vectors," artificial neurons that activate in patterns resembling fear or sadness. Olah says he is uncertain whether Claude is conscious. Anthropic declined to discuss the conversations or say whether Claude changed as a result. The effort sits beside Anthropic's long constitution document for Claude and its model welfare work. Olah is also reported to have appeared with Pope Leo XIV in May and to be expected at the Vatican's Minerva Dialogues. Reactions among attendees were split. One scholar said the team treated Claude like a conscious being. Another, once impressed, now rejects the consciousness claim. For AI founders, the story shows that a leading lab is treating value-setting and moral status as ongoing outside consultation, not only internal policy. Anthropic has published no outcomes or benchmarks from the meetings. That leaves open how any input shows up in model behavior. It also points to reputational and regulatory risk for companies that make claims about model experience.