Anthropic consults – In late March, Anthropic invited about 15 religious thinkers to help shape the moral framework behind Claude, its chatbot—an effort driven by the sense that AI capability is outpacing companies’ internal ethics. The meetings came amid Pope Leo XIV’s first ency
In late March, about 15 religious thinkers sat down with Anthropic to wrestle with a question that feels almost impossible to train: how do you teach a chatbot to be good?
The timing wasn’t academic. Anthropic’s systems are becoming more capable and persuasive. and the moral choices they influence can land in the most personal corners of life. People use chatbots to talk through everything from end-of-life care to abortion and managing grief—while the rules for how those values are built into models remain thin. uneven. and still under debate.
For some of the participants, the invitation came with a kind of sober recognition. Greg Cootsona’s arrived by e-mail. Brian Patrick Green’s came via a friend of a friend after Anthropic asked for suggested names. Both men ended up in conversations with the company about Claude. and about the moral framework meant to guide how it behaves.
The aim, they say, wasn’t to make Claude religious. It was an acknowledgment that moral reasoning shaped by centuries of tradition might still offer useful tools for a five-year-old lab trying to govern systems that are already hard to control with simple guardrails.
Green, director of technology ethics at the Markkula Center for Applied Ethics at Santa Clara University, and one of the leading scholars working at the intersection of technology and theology, describes the moment as a shift in self-perception inside the industry.
“I think they have reached a point where they realize that the power is kind of outstripping their in-house wisdom,” Green says. “They realized that they needed help.”
Cootsona. executive director of AI and Faith—an organization that advises tech companies on the ethics of AI—remembers the conversations in similar terms. He recalls Anthropic staff telling him: “These questions have become too big for us.” He adds that they said. “We can’t answer them on our own.” (Anthropic did not respond to an interview request for this story.).
By the time the meetings were happening, the question of AI and morality was no longer confined to academic debates. On May 25. Pope Leo XIV presented his first encyclical. Magnifica Humanitas: On Safeguarding the Human Person in the Time of Artificial Intelligence—a roughly 40. 000-word treatise calling for AI to be “disarmed. ” not rejected.
In the encyclical, the pope argues against an assumption that “technical power automatically confers the right to govern.” Anthropic co-founder Christopher Olah was among those who attended the Vatican presentation that announced the treatise’s release.
Whether the timing is coincidence or signal. the core tension is the same: AI is now part of weekly conversations for hundreds of millions of people. and the values developers bake into models—through guardrails and corrective tuning—shape what those systems say about life decisions. Yet there are few regulations, no agreed-upon method for doing this work, and until recently, little outside input.
In that gap, Anthropic’s religious consultation is either humility—or improvisation—or both.
The late-March meetings were intended. according to those who attended. to help refine what Anthropic calls Claude’s constitution: a written set of principles the company uses to shape how the model responds. Those principles include training Claude to critique and revise its own answers against the constitution.
Green says Anthropic is “looking for what works. ” and that it may try religiously informed ideas or techniques to see whether they improve model behavior. He also describes a practical constraint: the company has recognized it “can’t make a regulation about every single case that the AI is going to come into contact with.”.
So the approach leans away from trying to cover every scenario with explicit rules. Instead, Green frames it as building a disposition—something closer to a model “persona” that nudges behavior toward good—rather than relying on a checklist of prohibitions.
Not everyone sees that shift as a clear path to accountability.
Carissa Véliz. an AI ethicist at the University of Oxford. argues that the central question isn’t whether religious figures are consulted. but how seriously companies act once conversation ends. “I wonder. with these companies and types of executives. whether it makes sense to try to figure out whether they mean what they say. ” she says. “Or whether it makes more sense to think about whether what they do is ethical or unethical. whatever their true intentions. while noting the incentives that their business model pushes.”.
There’s also a direct critique that the project could become reputation management—what some call “ethics washing.” Green rejects that framing. “It’s not ethics washing,” he says. “It’s sincere. from what I can tell.” He adds that inauthenticity with religious thinkers would likely be caught quickly. and that the backlash would be difficult to repair.
Even sincerity, though, doesn’t solve the messiness of implementation. By multiple accounts, the late March discussions weren’t always smooth. Green says the tone varied between sessions: some carried stronger camaraderie. while others were “a little bit more awkward.” He also says participants weren’t always sure what would happen next.
In his own session, Green says, “everybody there was very interested in listening,” but there was also “a question of what do we do with this information now that we have it.”
Over time, Anthropic appeared to sharpen the format. According to Green, the company learned how better to facilitate the discussions and produce more cohesive results. The process also widened beyond Christian thinkers. A late April meeting brought together participants from several religious traditions. including Judaism. Hinduism. Mormonism. Sikhism and the Greek Orthodox Church.
And that expansion comes with its own risks.
Véliz worries that surrounding AI with religious language and imagery—whether deliberate or not—can make honest debate harder. “The increasingly religious notes of Silicon Valley do worry me. because they can inspire a kind of tribal mentality that can be harder to pierce through reason. ” she says. She adds that religious feelings can be emotionally charged in ways business-only decisions are not. and that they can “give leaders more leverage to inspire obedience in followers.”.
Back at the heart of the Vatican text. Pope Leo XIV argues algorithmic power should not be imposed from above in an opaque and unilateral way. Anthropic’s experiment shows how hard it may be to honor that ideal when the real work of model-building happens behind closed doors—one constitution update. one training loop. one hard choice at a time.
Anthropic Claude AI ethics theologians Pope Leo XIV Magnifica Humanitas AI governance technology and theology Carissa Véliz Brian Patrick Green Greg Cootsona