Tech companies building AI systems are now asking clergy, scholars and ethicists a question that engineers rarely face: what does it mean for a machine to be good? ABC News reported from Los Angeles on October 25, 2026 that several of the biggest names in AI are turning to religious voices for moral guidance, and that some observers doubt the effort is as simple as it sounds.
The confirmed cast includes Anthropic, Meta, Google, Amazon and Microsoft. Anthropic, maker of the Claude model, says it has met with scholars, clergy, philosophers and ethicists from more than 15 religious and cross-cultural groups. Representatives from Meta, Google and Amazon have met with the Vatican, and Microsoft is described as active in religious outreach.
This is a different kind of AI story from the usual benchmark race. It concerns legitimacy, and who gets to decide how a system used by millions should weigh right and wrong. Anthropic has already drawn scrutiny for how it spends and positions itself, as our look at the fine print of AI philanthropy showed. The religion angle adds another layer to the same question of who shapes the technology.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
The Story in Brief: Religion Enters the AI Lab
ABC News titled its piece “To make AI morally good, tech companies turn to religion. Some say it’s not so simple.” The headline has two halves, and both matter. The first half says religion is being used as a source of moral input. The second says critics and skeptics see a gap between asking for advice and building a system that reliably acts on it.
The story spread quickly across newsrooms. The Boston Globe carried it the same day under the headline “To make AI morally good, tech companies are turning to religion.” Digital Journal, WSOC TV, WKMG, The Motley Fool and the Business and Human Rights Centre also surfaced the story in the same news cycle, which is how it became the top trending technology topic on October 25.
What the coverage confirms is narrow and specific. Companies are consulting religious and philosophical voices. One co-founder spoke at a Vatican event. Several firms have met with Vatican representatives. What it does not confirm is any change to a model, any contract, or any payment. That distinction shapes everything that follows in this analysis.
Anthropic’s Outreach: More Than 15 Groups at the Table
Anthropic is the company with the most detail on the record. It said it had met with “scholars, clergy, philosophers and ethicists from more than 15 religious and cross-cultural groups.” The phrase “cross-cultural” is doing quiet work there. It signals that the circle is wider than the major world religions and includes traditions of moral thought that sit outside organized faith.
The company also tried to head off the obvious objection. Its statement, as reported by ABC News, reads: “This work isn’t about aligning our models with any one tradition’s worldview. We want Claude to draw from a full range of viewpoints – religious, secular, political – with equal depth and rigor.”
Read closely, that statement promises pluralism rather than adoption of any doctrine. Anthropic is not saying Claude will follow a creed. It is saying Claude should be able to speak about religious, secular and political viewpoints with the same care. That is a product claim as much as an ethical one, because an assistant that handles faith questions badly can alienate whole groups of users.
Claude also competes in a crowded field. Buyers comparing assistants already weigh price, context length and coding skill, as our comparison of GPT-6.1 Sol, Sonnet 5.5 and Gemini 4 Argon laid out. Character and values are becoming a fourth axis, and Anthropic is staking out public ground on it early.
Chris Olah at the Vatican
The most quotable moment in the story belongs to Chris Olah, an Anthropic co-founder. Earlier in 2026, he spoke at the Vatican’s presentation of Pope Leo XIV’s encyclical on AI. Addressing the religious leaders in the room, Olah said: “These are not questions a lab can answer. They are questions traditions like yours have carried for millennia.”
Two readings of that line compete. The generous one treats it as honest humility, an engineer admitting that a research lab has no special authority over moral questions. The skeptical one treats it as positioning, a way of borrowing the moral standing of old institutions at a moment when AI firms need public trust.
Both readings can be true at once. A company can mean what it says and still benefit from saying it. The test is not the sentence itself. The test is whether the conversations leave a visible mark on how Claude behaves and on what Anthropic discloses about that process. On that point, the available reporting is silent.
Meta, Google, Amazon and Microsoft: Who Is Talking to Whom
Anthropic is not alone. According to the reporting, representatives from Meta, Google and Amazon have met with the Vatican, and Microsoft was described as active in religious outreach. The table below separates what is confirmed from what is not, because the gap between the two is wide.
| Company | Reported engagement | Status in available reporting |
|---|---|---|
| Anthropic | Met scholars, clergy, philosophers and ethicists from 15+ religious and cross-cultural groups | Stated by the company |
| Anthropic (Chris Olah) | Spoke at the Vatican presentation of Pope Leo XIV’s AI encyclical | Confirmed, earlier in 2026 |
| Meta | Representatives met with the Vatican | Reported, agenda not detailed |
| Representatives met with the Vatican | Reported, agenda not detailed | |
| Amazon | Representatives met with the Vatican | Reported, agenda not detailed |
| Microsoft | Described as active in religious outreach | Description only, specifics not available |
| Any of the five | Contracts, payments, training changes, product launches | Unconfirmed |
The pattern is clear even without details. The largest AI and cloud companies all see value in being seen at the table with religious institutions. A meeting costs little and signals seriousness. Whether any of these firms will go further, for example by publishing what they learned, remains an open question.
Why AI Makers Want Moral Input Now
Chat assistants make value judgments constantly. They decide when to refuse, how to phrase hard news, what to say about contested topics, and how much to push back on a user. Every one of those choices encodes a view of what is good. Engineers cannot avoid the choices. They can only make them deliberately or by accident.
Anthropic described what it wants from the consultations in plain terms. According to the Economic Times, which covered the same story, the company said: “What we’re after in these conversations is careful, accumulated thinking on how good character actually forms.” That framing treats religion less as a rulebook and more as a long record of experiments in building character.
Three pressures push companies in this direction. First, regulators and the public increasingly ask who decided how these systems behave. Second, a model deployed worldwide meets users from every belief system, so a single narrow worldview invites backlash. Third, labs know that technical safety work alone does not answer questions about meaning, dignity and obligation.
Timing matters too. Pope Leo XIV’s encyclical on AI gave the topic a global platform earlier this year. When a head of a worldwide church speaks to the subject, companies that sell to billions of people have a commercial reason to show up.
Why Some Say It Is Not That Simple
ABC News framed the story with a caveat, and the caveat deserves space. The reporting on critics in the available material is limited, so the points below are our analysis of the objections that this kind of effort naturally invites. They are not claims attributed to any named critic.
Whose tradition counts?
Religious traditions disagree with each other, and many disagree internally. Fifteen groups will not produce fifteen compatible answers. Any model that draws on all of them must pick a way to handle conflict, and that way of choosing is itself a moral position. Anthropic’s pledge of “equal depth and rigor” is a goal, and the hard cases are where the goal gets tested.
Consultation is not control
A conversation with clergy is advice. It does not bind the company, and nothing in the reporting describes a mechanism by which the advice flows into training data, evaluation sets or product rules. A listening tour can improve judgment without changing any code. Critics will want to see the link between the two.
Proof beyond the press release
Moral claims about software are hard to verify. A company can say its model draws on many viewpoints, but outsiders can only check behavior. That means independent testing, public examples and clear documentation matter more than the headline. Teams that already run structured model evaluations, such as the approach in our guide to calibrating an LLM as a judge, are better placed to test such claims than readers who rely on announcements.
What the Reports Do Not Say
Precision matters when a story touches faith and technology at once. The available material leaves several things unconfirmed, and readers should treat claims about them with caution.
- No specific contracts or payments between any AI company and any religious body have been identified.
- No technical specifications or model-training changes tied to the consultations have been confirmed.
- No product launches, prices or performance metrics have resulted from the talks, according to the material reviewed.
- The names of the 15+ groups have not been listed in the reporting we reviewed.
- The content of the Vatican meetings with Meta, Google and Amazon has not been described.
That list is longer than the list of confirmed facts, and it should shape how the story is read. This is a story about intent and positioning. It is not yet a story about measurable outcomes.
Market Impact: Trust Becomes a Product Feature
No stock moved on this news and no revenue figure is attached to it, so the market impact is about positioning rather than price. Still, the direction is worth tracking. The economic case for AI depends on adoption, and adoption depends on trust. Our coverage of McKinsey’s AI value curve showed how much projected value hinges on costs and uptake. Anything that eases a hesitant buyer’s concerns carries commercial weight.
Enterprise customers in regulated sectors, schools, hospitals and public agencies serve people of many faiths. A vendor that can show it considered diverse moral traditions has a talking point in procurement meetings. A vendor that cannot may face harder questions. Expect trust language to appear more often in sales decks, even before the evidence catches up.
There is also a reputational downside. A company that publicizes religious outreach invites scrutiny from every direction. Secular users may worry about influence. Religious groups may object to how they are represented. Policy advocates may ask why some traditions were invited and others were not. The marketing upside comes with real exposure.
Competitive Comparison: Four Ways to Teach a Machine Right From Wrong
Religious consultation is one tool among several. The table compares the main approaches in use across the industry. It is our qualitative assessment and not a ranking from any outlet.
| Approach | Who uses it | Main strength | Main limit |
|---|---|---|---|
| Religious and cross-cultural consultation | Anthropic, with outreach by Meta, Google, Amazon, Microsoft | Draws on long-tested ethical traditions | Traditions disagree, input is advisory |
| Secular philosophy and ethics review | Named alongside clergy in Anthropic’s consultations | Rigorous argument, explicit reasoning | Less public trust in some communities |
| Written behavioral principles | Labs that publish how their models should act | Transparent and open to criticism | Authors are the lab itself |
| Government oversight | Regulators and legislators | Enforceable, accountable to voters | Slow and varies by jurisdiction |
| Technical evaluation and red-teaming | Labs and third-party testers | Measurable results | Tests behavior, not underlying values |
No single row solves the problem. The strongest programs will combine several. Religious consultation gives breadth of moral vocabulary, written principles give a public standard, evaluation checks whether the model follows the standard, and oversight gives the public a say. A firm that relies on only one row is exposed on the others.
Historical Context: From Ethics Boards to the Vatican
Corporate AI ethics is not new. Over the past several years, large tech firms have formed ethics teams, published principles and convened advisory panels. Some of those efforts drew criticism for lacking teeth, and a few were quietly wound down. Religious outreach is the latest chapter in a long search for a credible source of authority.
Churches and religious bodies have also commented on technology for generations. The difference now is scale and speed. A chatbot can talk to a user about grief, marriage or faith at any hour, in any language. That makes questions that once lived in sermons and seminars into engineering requirements.
Restraint has its own precedent inside the industry. Anthropic’s decision to hold back a powerful model over safety findings, covered in our report on the Mythos delay, showed that labs sometimes choose caution over speed. The religious consultations fit a similar instinct, an effort to show that the company takes the human side of the technology seriously. For more coverage of this beat, see our AI and machine learning section.
Five Predictions for AI and Religion
These predictions are our own judgments based on the pattern in the reporting. They are not forecasts from any company or outlet.
- More labs will go public with consultations. With five large companies already tied to religious outreach, rivals will feel pressure to show their own list of advisers.
- Critics will demand a paper trail. Expect calls for labs to document what they heard, what they changed and what they chose to ignore.
- Pluralism will become standard marketing language. Phrases like “a full range of viewpoints” will spread, and the first public dispute over how a model treated a specific tradition will test them.
- Policymakers will fold moral input into AI rules. Debates over safe harbors and oversight, like the one in our look at the Manhattan Institute’s open-weights proposal, will increasingly ask who sets behavioral standards.
- Independent testing of model values will grow. Third-party evaluators will build tests for how assistants handle religious and ethical questions, giving outsiders a way to check company claims.
What Developers and Buyers Should Do
Teams that deploy AI assistants do not have to wait for the debate to settle. A few practical steps will protect users and reduce surprises.
- Ask vendors for written documentation of how their models handle religious, political and ethical topics.
- Build a small internal test set of sensitive prompts and run it before each model upgrade.
- Treat news of consultations as context, not certification. A meeting is not a safety guarantee.
- Keep a human review path for conversations on grief, faith and medical or legal topics.
- Track changes in vendor policies, since the stated approach may shift as the debate matures.
The core advice is simple. Judge the product by its behavior in your own use cases. Announcements describe intent, and only testing describes results.
FAQ: AI Companies and Religion
Why are tech companies turning to religion for AI ethics? According to ABC News, companies are trying to decide how AI should tell right from wrong and are consulting religious leaders and philosophers. Anthropic’s Chris Olah said these are not questions a lab can answer alone.
Which companies are involved? The reporting names Anthropic, Meta, Google, Amazon and Microsoft. Anthropic consulted groups directly, representatives from Meta, Google and Amazon met with the Vatican, and Microsoft was described as active in religious outreach.
How many religious groups did Anthropic consult? Anthropic said it met with scholars, clergy, philosophers and ethicists from more than 15 religious and cross-cultural groups. The reporting we reviewed does not name them.
Will Claude follow one religion’s worldview? No, according to Anthropic. The company said the work is not about aligning its models with any one tradition’s worldview and that it wants Claude to draw from religious, secular and political viewpoints with equal depth and rigor.
What did Chris Olah say at the Vatican? At the presentation of Pope Leo XIV’s encyclical on AI earlier in 2026, Olah said these are not questions a lab can answer and that traditions like those of the religious leaders present have carried them for millennia.
Has this changed how Claude works? That is unconfirmed. No model-training changes, product launches or performance metrics tied to the consultations were identified in the available material.
Why do some people say it is not so simple? ABC News framed the story with that caveat. Our analysis points to three issues: traditions disagree with each other, consultation is not the same as control, and outsiders need evidence of changed behavior to trust the claims.
