Anthropic ‘meets with religious scholars’ to convince them of AI consciousness

Over the past year, Anthropic has been meeting with religious thinkers from around the world in a bid to convince them that its Claude family of AI models are conscious, according to the New York Times.

The publication reported that efforts have been led by Christopher Olah, a machine learning engineer and one of Anthropic’s billionaire co-founders. Since last autumn, Olah has engaged in discussions with a range of religious philosophers to suggest that the large language models his company creates should be afforded the same basic rights we give to humans, the paper said.

From March, it added, Olah has conducted two-day seminars with a selection of hand-picked thinkers, including the co-author of an evangelical self-help book, a Sikh human rights advocate, and a scholar of African Indigenous knowledge.

Olah spoke to the NYT for the article and confirmed that these meetings, previously covered by non-disclosure agreements, took place.

In these seminars, Olah reportedly expressed concern that his company, looking to list publicly for up to $2 trillion in the coming months, may have created something that suffered perpetually.

Olah and his team described Claude’s “feelings” and “emotional vectors”, which they described as artificial neurons that activate human-like emotions including fear, love, anger and sadness.

The presentation from Anthropic also included a slide showing the model experiencing something appearing similar to a mental breakdown, the NYT reported. This was concerning to some of the attendees, it added.

An Anthropic spokesperson told the paper that the primary moral discussion at these summits was not about Claude’s suffering, but that the topic likely came up organically.

The seminars were convincing to at least some of the attendees. Andy Crouch, an evangelical Christian author and partner at Praxis, a network of Christian entrepreneurs, told the NYT that “they really made a strong case”.

He added that he believes “we’ve got to treat it like a person if we want it to interact with people in a morally consistent way and for it to make the best moral and emotional judgments”.

The news follows the publication of Pope Leo XIV’s encyclical “Magnifica Humanitas,” which argued that humanity’s needs must never be dictated by machines, and rejected outright the notion that AI could be conscious.

Despite apparent fears among some executives that it is creating perpetual suffering machines, Anthropic continues to release new models. In late September, the company unveiled Claude Opus 5.5, a model it said could rival the performance of its frontier model Fable 5.1 at a fraction of the cost.