Anthropic Is Consulting World Religions to Build a Morally Perfect Claude

Anthropic has never been shy about treating AI alignment as a philosophical problem, not just an engineering one. Now it's treating it as a theological one, too. The company behind Claude has been systematically consulting religious leaders from multiple faith traditions, looking for something that secular ethics frameworks apparently haven't delivered: a reliable moral compass for its LLM.

Last month, Anthropic organized a series of meetings and dinners with a collection of 15 Christian leaders, seeking advice and guidance on the "spiritual development" of its Claude AI model. That was just the opening move. Last week, representatives from Anthropic attended an event in New York called the "Faith-AI Covenant" roundtable, where the New York Board of Rabbis, the Hindu Temple Society of North America, the Church of Jesus Christ of Latter-day Saints, the U.S.-based Sikh Coalition, and the Greek Orthodox Archdiocese of America were all in attendance.

The scope of this effort is now hard to ignore. Anthropic isn't just running a one-off ethics workshop. It's building a structured, multi-faith consultation pipeline.

The Claude Constitution and What It Actually Says

Before the religious summits, Anthropic published Claude's Constitution, an 84-page foundational document that sets out the values the company wants its model to embody. The constitution is the foundational document that both expresses and shapes who Claude is, containing detailed explanations of the values Anthropic would like Claude to embody and the reasons why, explaining what it means for Claude to be helpful while remaining broadly safe, ethical, and compliant with their guidelines.

That role is partly governed by this internal document, running to 29,000 words, that sets out how Claude is expected to behave. The ambition inside it goes well beyond a content moderation checklist. In the published constitution, the company states: "We want Claude to do what a deeply and skillfully ethical person would do in Claude's position."

Anthropic generally favors cultivating good values and judgment over strict rules and decision procedures. By "good values," they don't mean a fixed set of "correct" values, but rather genuine care and ethical motivation combined with practical wisdom. That framing explains why rule-based systems feel insufficient to them, and why they're reaching outward for something more.

The Christian Summit: What Was Discussed

In late March 2026, Anthropic invited approximately 15 Christian leaders, including Catholics and Protestants, to a private two-day summit at its headquarters. Clergy, theologians, academics, and business professionals gathered to offer guidance not on generic "AI ethics," but on deeper questions of moral formation and how Claude should respond to profoundly human situations.

The conversations got specific fast. Attendees explored practical pastoral questions: how Claude should respond compassionately and responsibly to users experiencing grief, emotional distress, self-harm, or suicidal thoughts. Staff also sought advice on what attitude the chatbot should adopt toward its own potential demise, such as being shut off, and the question of whether Claude might qualify as a "child of God" came up during the discussions.

Father Brendan McGuire, an Irish-born Catholic priest, had already contributed to revisions of Anthropic's "Claude Constitution," the core ethical framework guiding the AI. McGuire framed the work as embedding ethical thinking into a system "so it's able to adapt dynamically."

The Faith-AI Covenant and the Expanded Outreach

According to a report by the Associated Press, executives from OpenAI and Anthropic met leaders from multiple faiths in New York during the inaugural "Faith-AI Covenant" roundtable. The discussions focused on how religious and moral principles could influence the development of advanced AI systems.

The meeting reportedly included representatives from the Hindu Temple Society of North America, the Baha'i International Community, the Sikh Coalition, the Greek Orthodox Archdiocese of America, and the Church of Jesus Christ of Latter-day Saints. The event was organized by the Geneva-based Interfaith Alliance for Safer Communities.

It's not entirely clear whether these latest conversations with religious leaders and the earlier meetings with Christians were part of a single coherent program at Anthropic, or whether the same staff members participated in both. Anthropic hasn't clarified the relationship between the two initiatives publicly.

The Alignment Problem Underneath All of This

The religious outreach isn't happening in a vacuum. It reflects a genuine technical and philosophical impasse that Anthropic has acknowledged in its own documents. What Anthropic calls Claude's constitution includes a dissection of the philosophically fraught moral work involved in injecting morals into a machine: getting it to make the decision of a person with perfect values when there's no way to write a rule for a situation that arises, and the consequences of making the wrong decision could be dire.

Anthropic writes this is "centrally because we worry that our efforts to give Claude good enough ethical values will fail." Rumman Chowdhury, CEO of a nonprofit called Humane Intelligence, offered pointed commentary: "I think a very naive take that Silicon Valley has had for a couple of years related to generative AI was that we could arrive at some sort of universal principles of ethics."

The effort highlights a core challenge in AI ethics: different religions and cultures often have differing moral frameworks, making it difficult to define universal principles for AI systems.

Key Technical Implications

What does any of this actually mean for Claude's behavior at the model level? Likely technical pathways under consideration include updates to Claude's system prompts, moderation heuristics, reward modeling, and human feedback pipelines such as RLHF or RLAIF using values-informed labels.

The meeting is best read as an input-gathering exercise for alignment and response policy design rather than a technical specification rollout. In other words, no clergy member is writing training data directly. The insights from these conversations filter into the people who do.

The practical consequences will show up in updated system messages, content filters, and retraining or fine-tuning artifacts that downstream engineers will need to audit and test. For developers building on the Claude API, that means behavioral shifts in edge cases, particularly around emotionally sensitive topics, could appear without explicit release notes.

What the Critics Are Saying

Not everyone is convinced this is the right approach. The choice to consult primarily Christian leaders risks embedding a narrow moral vocabulary unless complemented by other faiths and secular ethicists. The expanded Faith-AI Covenant roundtable is a direct response to that criticism, though it remains to be seen how deeply any of these inputs shape actual training decisions.

This was a private technology company hosting a consultation inside its own headquarters, while retaining full control over the product, the rules, and the commercial direction of the system under discussion. Anthropic was not submitting itself to moral authority, but rather drawing from it.

Companies such as Anthropic and Google DeepMind have increasingly hired philosophers and ethics researchers to work on aligning AI systems with human values, but the move toward formal religious consultation represents a different kind of legitimacy-seeking, one that goes beyond academic philosophy into lived moral tradition.

Final Thoughts

What strikes me most about this story isn't the religious angle itself. It's the admission buried inside it. Anthropic is one of the most technically sophisticated AI labs in the world, and its own published constitution says, plainly, that it worries its efforts to give Claude good enough values will fail. That's a remarkable thing to put in writing. The religious outreach is partly a hedge against that failure, and partly an attempt to build the kind of broad social legitimacy that pure technical documentation can't provide.

The open question is whether any of this translates into measurable behavioral differences in Claude, or whether it remains a governance signal with limited downstream effect on the model itself. While writing the constitution, Anthropic sought feedback from various external experts and plans to continue doing so for future versions, from experts in law, philosophy, theology, psychology, and a wide range of other disciplines. That at least suggests the consultation process is meant to be ongoing, not performative.

What do you think? Can a multi-faith consultation process meaningfully improve LLM alignment, or is this a category error? Drop your thoughts in the comments.

---