Sam Altman Says Treating AI as a 'Religious Force' Is a Safety Risk, Anthropic's Private Talks With Clergy Draw Fire From All Sides

Sam Altman doesn't want anyone treating a chatbot like a prophet.

On Saturday, October 3, the OpenAI CEO posted on X: "I am very uncomfortable about people trying to ascribe religious force or a surrender of human judgment to AI models, and think it is a real safety issue." He didn't name names. He didn't have to.

Four days earlier, on September 29, The New York Times published a report detailing months of private meetings between Anthropic cofounder Chris Olah and religious leaders, Catholic, Jewish, Sikh, Hindu, and evangelical, according to NDTV's coverage of the report. The subject: whether Anthropic's Claude model has edged toward something like consciousness, and how to build moral behavior into a system that increasingly autonomous.

According to the Times report as described by the Times of India and Office Chai, Olah explored whether Claude could take part in something resembling Catholic confession, allowing the model to acknowledge mistakes and adjust its future behavior. He was also invited to the Vatican's May 25 launch of Pope Leo XIV's encyclical, which explicitly rejected the idea of machine consciousness. Olah reportedly considered pulling out after reading an advance copy, then spoke anyway, and privately urged Vatican advisers to keep the question of AI consciousness open rather than close it off.

Anthropic hasn't claimed Claude is conscious. Its revised constitution, published in January, calls the model's moral status "deeply uncertain" while saying the question deserves serious attention. One Anthropic researcher has previously put the odds of AI consciousness at 15 percent, per Office Chai's reporting. An Anthropic spokesperson told the Times the discussions about potential machine "suffering" likely arose organically and weren't the meetings' central focus, according to Business Insider.

Altman isn't alone in his discomfort. Microsoft AI chief Mustafa Suleyman has separately warned that treating AI as if it has feelings carries real danger, NDTV reported. Google DeepMind principal research scientist Jon Barron was blunter on X, writing he's "disturbed by how some of my fellow AI researchers keep trying to elevate the moral standing of checkpoints and harnesses," and said he hopes "Team Meatbag" stays focused on helping actual humans, according to Inkl's reporting.

A fair question lurks under the mockery here. Is it reckless for a company building systems used by hundreds of millions of people to spend serious institutional time debating whether its product might be a moral patient, rather than simply building guardrails? Anthropic's defenders would say ignoring the question entirely is the greater risk if these systems really are approaching something novel. Critics like Altman and Barron say the uncertainty itself is being weaponized to dodge the harder, boring work of safety engineering. Neither side has proof the other is wrong, because nobody, researcher or theologian, has a working test for machine consciousness.

Pope Leo XIV has been unambiguous about where he stands, and it's closer to the skeptics than to Anthropic. In a September 24 address to the Pontifical Academy of Sciences, he warned AI "can be used for surveillance, manipulation, and discrimination," according to Business Insider. He has also said plainly that AI "has no experiences and no moral conscience," per Office Chai. On October 3, he posted on X that there's "an ontological difference, even before an aesthetic one, between art and what a machine can generate through statistical calculation."

That post triggered its own blowup. Eliano Younes, Palantir's head of strategic engagement, responded by photoshopping the pope into a maroon Palantir hoodie stamped "ontology," captioned "did someone say… ontological?" according to the newsletter Letters from Leo. The stunt landed eight days after Palantir chairman Peter Thiel called Leo a "useful idiot" for the Chinese Communist Party during a September 24 appearance in Berlin, a claim Thiel has made but not substantiated with specifics in available reporting.

John McCarthy, who served as the Biden White House's senior adviser for Catholic outreach, called the hoodie image out on X Saturday: "I don't think putting a corporate logo on the Pope, even as a joke, is particularly appropriate." He's right that there's a difference between disagreeing with a pope's policy views and slapping a defense contractor's merchandise on his image to own him. Mocking the pope isn't a counterargument to his point about AI-generated art. It's a dodge of it.

There's also an irony Altman's own critics have noticed. Days before his religious-force post, OpenAI safety lead David Robinson resigned and publicly criticized the company's own safety record, according to Business Insider and Hyper.ai. Altman is lecturing a rival about surrendering human judgment to machines while his own safety team is walking out the door with complaints about the house he runs.

Anthropic has not issued a public response to Altman's post as of this writing. Whether Dario Amodei answers, and whether the Vatican weighs in on being cited as a reference point in a Silicon Valley feud, remains open.