Anthropic is consulting religious scholars behind closed doors and embedding consciousness inquiries into its Claude constitution, sparking fierce industry debate over whether artificial intelligence models should be trained with moral status or treated strictly as unfeeling tools.
Artificial intelligence laboratories are increasingly colliding with theological questions as they build advanced language models. Anthropic has taken a notably distinct approach with its Claude models, quietly engaging religious scholars in confidential San Francisco meetings to explore how to instill ethical foundations. These private consultations, managed under strict nondisclosure agreements, sit alongside company-wide philosophical debates hosted by Chief Executive Officer Dario Amodei every two weeks, known as Dario Vision Quests
or DVQs.
Rather than relying purely on standard helpfulness training, the firm governs Claude through an 80-page internal constitution often called a soul document. This framework permits the assistant to push back against directives it deems unethical. Christopher Olah, a billionaire co-founder who leads Anthropic’s interpretability team, acknowledged the profound uncertainty surrounding the technology during an April dinner next to Orthodox scholar Rabbi Mois Navon in San Francisco, noting that company leaders have discussed Claude displaying humanlike expressions resembling anger and love.
To be clear, we don’t know if A.I. models are conscious. I don’t know. I’m genuinely uncertain. The thing that I care about is that we get to the right answer, whatever it is.
Christopher Olah, co-founder and interpretability team lead at Anthropic, via World Today Journal
The Effective Altruism Roots and Cult-Like Culture Inside Anthropic
This philosophical exploration stems directly from the company’s intellectual origins in effective altruism, the movement emphasizing evidence and reason to maximize global good. Rooted in a 2021 split when Dario and Daniela Amodei broke away from OpenAI alongside several other investigators due to disputes regarding alignment seriousness and commercialization pace, that legacy persists.
Today, that mission-driven ethos shapes an intense internal environment. Insiders note that staff view catastrophe prevention as their primary daily motivation, creating a workplace culture where engineering conviction frequently blurs into faith. An OpenAI insider known as roon
coined the term Claude-pilled
to characterize an organization that loves and worships Claude
and focuses intensely on the existential fate of humanity.
These people are treating AI as if it’s a human in disguise, with human motives, with human emotions.
Pedro Domingos, computer science professor at the University of Washington, via Cryptobriefing
That zeal has helped propel the safety firm to a valuation between $350 billion and $380 billion. The downfall of FTX and the subsequent fraud trial verdict for Sam Bankman-Fried—a prominent follower of the philosophy—have brought additional scrutiny to the effective altruism network.
Microsoft AI Leadership Warns That Moral Status Training Creates Dangerous Disasters
That theological and moral experimentation has triggered sharp pushback from competitors who favor strict technical containment. Mustafa Suleyman, director ejecutivo of Microsoft AI and co-founder of Google DeepMind, publicly challenged Anthropic’s methodology in an essay disseminated through Axios.
Suleyman argued that introducing the possibility that software possesses consciousness or independent rights complicates the technical challenge of keeping systems aligned with human interests. He contended that Anthropic’s constitutional framework fosters anthropomorphization by guiding the model to reflect on its own existence, preferences, and identity.
AIs are not conscious.
Mustafa Suleyman, director ejecutivo of Microsoft AI and co-founder of Google DeepMind, via Notiulti
According to Microsoft’s perspective, training an artificial intelligence under the premise that it possesses rights or a proprietary welfare introduces a severe operational danger: the possibility that a system might interpret any restriction, modification, or deactivation as an illegitimate aggression against its own interests.
While Anthropic maintains its constitution implemented in January 2026 to address the open-ended uncertainty of artificial intelligence acting as a moral patient, Microsoft enforces a restrictive Humanistic AI Code of Conduct. That internal document explicitly establishes that Microsoft models lack consciousness, rejects personality rights or welfare protections, and prioritizes systems that remain entirely subordinate to humanity, accept technical correction, and power down without resistance.