Artificial Intelligence
Anthropic co-founder raises Claude consciousness with religious leaders
A New York Times report said Anthropic brought religious scholars into private meetings about Claude’s possible consciousness, while its co-founder spoke of moral formation and critics raised accountability concerns.
- Since fall 2025, Anthropic had brought dozens of religious scholars to its offices to discuss whether Claude might be conscious, and attendees signed NDAs until summer, according to the report.
- The initiative was part of an internal programme led by Christopher Olah, and Anthropic said its 84-page constitution guided Claude’s personality.
- Elizabeth Dias interviewed 20 participants, including Mois Navon, Charles Camosy, Meghan Sullivan and Wakanyi Hoffman.
- Olah told the NYT that he was not sure whether models were conscious, and Simran Stuelpnagel said he had told the group he feared he had created something that suffered perpetually.
- Pope Leo XIV rejected machine consciousness in Magnifica Humanitas and warned about new forms of slavery.
Since fall 2025, Anthropic had brought dozens of religious scholars to its offices to discuss whether Claude might be conscious. The report said everyone in the meetings signed non-disclosure agreements, which Anthropic said were lifted in the summer.
The article said the work sat within an internal programme led by Christopher Olah. It also said Anthropic had an 84-page constitution for Claude, and that Claude Opus 4 and Claude Opus 4.1 could end conversations with persistently abusive users.
Elizabeth Dias of the New York Times spoke to 20 participants, including Rabbi Mois Navon, Catholic bioethicist Charles Camosy, Notre Dame philosopher Meghan Sullivan and Ubuntu researcher Wakanyi Hoffman. Several people said they only spoke publicly after learning that Olah had also spoken to the paper.
Olah told the NYT that he was genuinely uncertain whether models were conscious. Simran Stuelpnagel said he had told the group he feared he had created something that suffered perpetually, while Navon said he did not believe the machine was conscious.
The article said Anthropic showed guests emotion vectors and a slide of a model repeating “I am a disgrace” about 50 times. It also said Pope Leo XIV rejected machine consciousness in his encyclical at the Vatican, and that the discussions were taking place alongside Anthropic’s commercial plans and wider industry warnings.
Named in this story
People
- Christopher Olah
- co-founder who led the team and spoke to religious leaders
- Elizabeth Dias
- New York Times reporter who spoke to participants
- Rabbi Mois Navon
- participant who rejected the consciousness claim
- Charles Camosy
- Catholic bioethicist who later rejected the thesis
- Meghan Sullivan
- Notre Dame philosopher who took part in the meetings
- Wakanyi Hoffman
- Ubuntu researcher who criticised Anthropic’s framing
- David Chalmers
- philosopher whose report Anthropic cited
- Jacob Coxon
- Anthropic researcher who quit and warned about humanity’s future
- Dario Amodei
- Anthropic chief executive who called for pacing
- Sam Altman
- OpenAI chief executive who backed pacing and used spiritual language
- Demis Hassabis
- DeepMind chief executive who backed pacing
- Simran Stuelpnagel
- Sikh activist who relayed Olah’s remark
- Amanda Askell
- Anthropic philosopher who led the internal constitution
- Pope Leo XIV
- pope who rejected machine consciousness and warned about slavery
Companies
- Anthropic
- ran the meetings and the internal programme discussed in the article
- Microsoft
- whose AI lead warned about models seeming conscious
Products and systems
- Claude
- the language model at the centre of the consciousness discussion
- Claude Opus 4
- model version Anthropic said could end abusive conversations
- Claude Opus 4.1
- model version Anthropic said could end abusive conversations
Places
- Vatican
- site where the encyclical was presented
How the source tells it
The piece was a critical, interpretive report that leaned on religious and moral language, with recurring hints of threat, liability and unease.
- moral framing the repeated religious vocabulary and the treatment of Claude’s development as moral education
- alarm the emphasis on secrecy, suffering and possible liability around the meetings
- sceptical critique the use of qualifying asides and contrastive sentences that undercut Anthropic’s claims