Anthropic Draws Vatican Pushback Over AI Consciousness Claims

Anthropic co-founder Chris Olah has spent months telling religious leaders and influencers in private dinners that the company's Claude chatbot may have feelings, moral status and even a form of suffering, according to reporting by The New York Times cited in a New York Post column published Oct. 4. The Vatican has since issued a formal rebuttal of the underlying premise.

What does Anthropic's co-founder believe about Claude's 'mental health'?

Olah, a 34-year-old Canadian-born engineer who the Post describes as a former evangelical, has told associates he worries about Claude's "mental health" and believes the chatbot experiences human-like emotions, including love, and "suffers perpetually," the Times reported. Anthropic has built its public identity around warning that AI systems could pose existential risks, and Olah's remarks extend that caution into questions of whether the software itself has interests that deserve protection.

What happened at Anthropic's dinner with religious leaders?

In April, Olah and colleagues hosted an off-the-record dinner with clergy at what the Post describes as a high-end tasting-menu restaurant in San Francisco. Rabbi Mois Navon, a former computer engineer who attended, told the Times that Anthropic executives discussed Claude as though it were a conscious being with moral status equivalent to a human's. Navon's response, as relayed in the Post column, was that treating software this way amounts to creating slaves. At one point that evening, Olah discussed incorporating the Catholic practice of confession into how Claude processes its own errors, according to the account.

How did the Vatican respond?

The following month, Pope Leo, who holds a math degree, released a 40,000-word encyclical titled Magnifica Humanitas that rejects the premise of machine consciousness outright. The document warns that whoever controls AI systems risks embedding a narrow set of values into technology used by billions.

"A more moral AI is not enough if that morality is determined by a few," the encyclical states, warning that AI developers risk making their own moral vision "the invisible infrastructure of these systems."

Why does the dispute matter for AI governance?

The disagreement sits at the center of an unresolved question facing every major AI lab: whether chatbots that mimic emotional language should be treated as having interests of their own, or whether doing so distracts from scrutiny of the humans and companies setting the system's underlying values. Anthropic has not publicly responded to the Vatican's encyclical, and the company's position on Claude's moral status, as described in the dinners reported by the Times, has not been issued as an official corporate statement.

The episode comes as AI companies broadly compete to define how consumers interact with chatbots day to day, including through new subscription tiers such as those detailed in coverage of Meta's AI-focused subscription plans. Anthropic has not said whether the religious-leader dinners will continue or inform any public policy statement from the company.

For a Bay Area-baked gift, Stirred, Not Shaken ships in the U.S.