Anthropic, the company behind the AI model Claude, has spent much of this year convening religious leaders to help shape the moral reasoning built into its system. The New York Times reported the story on September 29, 2026, in an article by Elizabeth Dias. What does this change for the millions of people who use Claude? Potentially a great deal — and almost none of it is verifiable from outside.

The reported facts are these:

• A first meeting this spring gathered roughly fifteen Catholic and Protestant leaders.
• Later rounds added participants from Judaism, Hinduism, the Latter-day Saints, Sikhism, and Greek Orthodoxy.
• Participants recall co-founder Chris Olah telling them these questions had become too big for the company to answer alone.
• The subject of refinement was Claude's "constitution" — the long document of principles that governs how the model behaves.

That constitution's January revision states openly that Claude's moral status is deeply uncertain. Anthropic deserves credit for writing that down; every promise in this story, however, comes with an asterisk.

What was happening at the same time?

The consultations ran in parallel with the preparation of Pope Leo XIV's encyclical Magnifica Humanitas, presented on May 25, which calls for AI to be "disarmed" rather than rejected. According to the Times, Olah later pressed Vatican advisers to take the possibility of machine consciousness seriously. A Silicon Valley researcher lobbying the Vatican on the nature of the soul is a sentence that should slow any reader down.

Dennis P. Petri, International Director of the International Institute for Religious Freedom, whose analysis of the report was republished from the Five4Faith Substack, concedes the point in Anthropic's favor. Admitting that ethical questions are too big for engineers, and turning to the traditions in which most of humanity actually reasons morally, is more honest than presenting values as neutral configuration settings. He is right. But he then asks the three questions any investigator should ask of a consultation: who was in the room, how they were chosen, and on what terms.

Who was in the room?

Here the answers are thin. The participants were invited by the company. Their deliberations were confidential. Swami Sarvapriyananda, one of the participants, told an audience he could not share details because he had signed a non-disclosure agreement. He also said he was impressed by what he saw — a claim we have no reason to doubt and no way to test.

No report mentions Muslim or Buddhist participants. Perhaps they were present. We cannot know, which Petri argues is precisely the problem: the moral formation of a system used by millions is being informed by a handpicked group whose discussions the public cannot read. The Oxford ethicist Carissa Véliz has asked whether the consultation amounts to "ethics washing" — consultation performed for appearance rather than substance. The fair answer, Petri notes, is transparency, not reassurance.

Is Claude conscious?

Olah's reported answer: he is genuinely uncertain. The religious leaders left divided. Rabbi Mois Navon observed staff relating to the model as if it were a conscious being. The Catholic bioethicist Charles Camosy concluded it is not.

The sharpest critique came from a competitor. Mustafa Suleyman, who leads Microsoft's AI work, argues Anthropic has built a circle: the model trains on a constitution discussing its possible moral status, reflects those ideas back, and the reflection is read as evidence. "Simulating consciousness is not the same as having it," he wrote. "A computer model of a hurricane will not blow your roof off."

Why does this matter commercially?

Claims about moral status carry political and legal consequences. Once a company speaks of its product's welfare, the vocabulary of dignity and rights follows. Petri, who documents religious-freedom violations against people whose rights remain unprotected, is in no hurry to extend that vocabulary to software.

His two rules for AI use apply here in reverse. Declare it: if religious traditions shape a machine millions consult, the public deserves to know who said what. And do not outsource: the deepest risk is not that Claude might have a conscience, but that the rest of us stop exercising ours. Uncertainty, on Anthropic's own admission, is a reason for caution in both directions — not a license to build the conclusion into the training data.