Sam Altman would like people to stop treating chatbots as if they might have souls. The OpenAI CEO said on Saturday that he is “very uncomfortable” with the notion of attributing some form of religious power to AI models—a comment Axios characterized as a veiled shot at Anthropic, the rival lab that has been exploring what it means for an AI to have a moral core.

Altman shared the remarks in a post on X on October 3, 2026. He framed the attribution of religious status to AI not as a matter of personal taste but as a genuine safety concern—placing a dispute that might sound like theology squarely inside the AI industry’s running argument over how much autonomy models should be granted. He did not name Anthropic directly, and he did not really need to: the timing of his post lined up neatly with a New York Times investigation published September 29–30, 2026.

That report detailed an initiative led by Anthropic co-founder Christopher Olah to bring together scholars from a range of religious traditions. Those gatherings reportedly began in Fall 2025, and their focus was on how morality might be built into AI systems, including thorny questions about consciousness and moral status. According to Axios, Anthropic has reportedly sought out scholars across faiths to help make the case for models that are moral and conscious.

Inside Anthropic’s ‘Soul Doc’

The centerpiece of Anthropic’s approach is a governing document for its Claude model. The constitution runs 84 pages and is known internally as the “Soul Doc.” It drew on input from philosophers and faith leaders. Documents of this kind serve as the reference point for how a model is meant to behave, which is why the language chosen for them carries practical weight beyond philosophy.

Altman’s objection appears to land on the step beyond that. Writing ethics into a system is one thing; suggesting the system itself might be a moral or conscious entity is another, and that is where he grows uneasy.

He is not alone in drawing that line. Pope Leo XIV has publicly rejected the idea that machines could be conscious—a notable position, given that religious voices are part of the very conversation Anthropic has tried to start.

An Old Rivalry With a New Front

Dario Amodei founded Anthropic in 2021 after leaving OpenAI, with the stated aim of putting AI safety and ethics at the center of the company’s work. Axios described the latest exchange as yet another point of division between Altman and his former colleague.

The split maps onto two distinct theories of safety. Altman’s camp treats humans as the responsible party, with AI as an instrument. Anthropic’s approach leans toward giving the model an internal moral compass of its own. Where those safeguards sit—in the people operating the system or in the system itself—has become one of the field’s most consequential design questions.

What This Means for the AI Industry

For businesses deciding which AI provider to work with, how a lab thinks about its models shapes how those models behave, refuse requests, and explain themselves to users.

There is also a public trust dimension. If people start attributing spiritual significance to chatbots, the potential for over-reliance grows. That appears to be the safety concern Altman is pointing at: users treating a model’s output as wisdom rather than as the product of training.

On the other hand, Anthropic’s outreach to religious and ethical traditions could appeal to customers and regulators who want evidence that AI developers are taking morality seriously.

Investors in the AI sector may want to watch how this divide affects partnerships and funding. A company’s stance on questions like model consciousness could start showing up in procurement decisions, policy debates, and reputational risk assessments. The open question is whether the disagreement stays a philosophical one or hardens into practical tests—in the products users touch and the rules regulators write.