Conference Proceedings · Volume 1

Deus Et Machina: An Interfaith Framework for Ensuring Divine Wisdom Guides Artificial Intelligence Development

Gaurav Rastogi
Hindu Spiritual Care Institute
Correspondence: vidya@ekrasworks.com

DOI: 10.68315/svpc26-ukj5retj
Publication date: November 19, 2026

This paper was presented at the Silicon Valley Peace Conference 2026 in Palo Alto, California, United States, held September 17–20, 2026, under the theme “AI’s Role in Community Development and Peacebuilding in Developing Nations,” and is published as part of the Proceedings of the Silicon Valley Peace Conference 2026.

Abstract

Artificial intelligence concentrates power. One religious tradition has answered with institutional seriousness but alone; the others are present in the work but have no way to speak as traditions, and a secular movement has filled the resulting space. The Hindu case shows the nature of the difficulty: Hindus are among the people building it, and what is missing is not participation, but a center authorized to speak. That is a problem of coordination rather than of recruitment, and it takes a different remedy. This paper proposes one, built as a wheel: a minimal charter inherited from what the traditions already hold in common, delegations drawn from institutions that exist already, and a small number of formed people placed inside the laboratories and the agencies, close enough to the work to be present when a model is cleared for release. Seven live questions in AI ethics are unasked rather than answered, and the implied answers converge on a short inventory of capacities that only formed people, not metrics, can supply. Two things are asked of the traditions: to hold a power to limits it did not set, and to say what it should be built for. No existing charter attempts the second. The hard objections are met directly. Deus Et Machina: God with the machine, against Ex-Deus Machina, the machine with the god taken out of it.

Keywords: interfaith dialogue, artificial intelligence ethics, Hindu philosophy, theological ethics, Deus Et Machina, concentration of power, rājadharma, sanātana dharma, digital colonialism

Introduction

On the morning of July 16, 1945, in the New Mexico desert outside Alamogordo, J. Robert Oppenheimer watched the first atomic device detonate at the Trinity site, a name he had chosen himself, from Donne’s “Batter my heart, three person’d God” (Atomic Heritage Foundation, n.d.). The fireball climbed into the sky, then the sound arrived, and finally the heat reached the bunker a few seconds after that. As the vision of the mushroom cloud rose before him, a line from the Bhagavad Gītā came to him: “Now I am become Death, the destroyer of worlds.” (The verse is Gītā 11.32, where the Sanskrit kāla means time; “Death” was Oppenheimer’s own rendering.) The line is oft quoted, but not very well understood. What it asks the reader to understand, we believe, is that Oppenheimer was not glorifying the weapon. He was registering, with atomic precision and solemnity, that human capability had now outrun the wisdom available to govern it, and that the distance between the two was where the danger lived.

He spent the rest of his life trying to close that distance by building institutions: chairing the advisory committee that argued, in 1949, against building the hydrogen bomb, pressing for international control of atomic energy, testifying and writing for general readers, organizing scientists to take up political questions they had been trained to avoid. This led to his security clearance being stripped in 1954, at a hearing that was less about secrets than about punishing the man who asked whether scientists should build everything they were able to build.

He failed; the hydrogen bomb was built anyway and first tested in 1952. The arms race ran on past anything he could do to slow it. It was not for lack of wisdom, for Oppenheimer had it himself, drawn in part from the Gītā he had learned Sanskrit to read. He is one of very few human beings to have stood beside an atomic detonation having read that book in the original language. What was missing was an institution with the ability to carry that wisdom and set it against the power. One deeply formed conscience is not an institution. Lives are short, but institutions live on.

Power without wisdom ultimately destroys. It is an old adage, but the destruction does not arrive at once, which is what makes it easy to wave off. The bomb worked, the war was over, and the destruction arrived structurally, in the form of an arms race that outlived everyone who started it. We are now living through a second crossing of that threshold, and the institutional gap Oppenheimer could not close has only grown wider.

That second crossing is the arrival of the Age of Intelligence (Altman, 2024) with the coming of Artificial Intelligence that can compute, think, and act. The bomb was built by one government in secret and used only twice; these systems are built by just a handful of firms in the open and used by billions of people every day. While the bomb threatened the body, this technology reaches the faculties by which a person understands anything at all. It has begun reshaping minds and societies before anyone has had a chance to observe the effect, let alone learn to wield these tremendous powers.

But now, no one stands where Oppenheimer stood, close enough to the work to see what it was and formed enough to ask what it was for. At that moment of reflection, he turned to one of the oldest continuous civilizations for words to express the enormity of what was unfolding. He turned to the Bhagavad Gītā, a conversation set at the momentary pause before a calamitous battle of annihilation, which has language for this experience that engineers might appreciate, even if they do not speak it.

Seated as on the Machine

The Gītā that guided Oppenheimer has a verse for exactly this problem. In Chapter 18, verse 61, Krishna tells Arjuna:

ईश्वरः सर्वभूतानां हृद्देशेऽर्जुन तिष्ठति। भ्रामयन्सर्वभूतानि यन्त्रारूढानि मायया॥

“The Supreme Lord is situated in everyone’s heart, O Arjuna, and is directing the wanderings of all living entities, who are seated as on a machine, made of the material energy” (Vyasa, n.d., 18.61).

Īśvara: Divine Intelligence that pervades all reality, coordinating existence toward truth and justice. This is an immanent presence rather than a distant clockmaker god, working through material mechanisms while transcending them. For AI ethics, it suggests that intelligence itself has a telos, a direction it inclines toward when ignorance and greed do not distort it. Whether our artificial systems incline that way or against it is not yet settled.

Hṛdaya: the heart-center where divine intelligence meets human consciousness, the place where wisdom and moral discernment reside. The heart in Hindu anthropology is epistemological as well as emotional, a way of knowing in which intellect and intuition are not at odds and analysis carries compassion within it. For AI development, the sophistication of the algorithm should be delineated from the quality of consciousness in the people who design and deploy it. An algorithm built by a designer whose heart is aligned with that intelligence will not produce the same thing as one built by a designer whose heart is captured by ego or fear or greed, even when the technical specifications are identical.

Poets, lovers, and artists all know the language of the heart, a knowing that arrives before knowledge, and across many traditions this is where the knowing self is placed. The Upaniṣads seat the self in the space within the heart (Chāndogya Upaniṣad 8.1.1; Kaṭha Upaniṣad 2.3.17), and the non-dual Shaivism of Kashmir goes further, making the Heart its name for consciousness itself (Muller-Ortega, 1989). The claim is not India’s alone. Sufi metaphysics treats the qalb as the organ of a perception that reason cannot perform (Corbin, 1969), and the Orthodox ascetic tradition describes the long work of bringing the intellect down into the heart (Ware, 1979). Blaise Pascal, who built one of the first mechanical calculators and knew exactly what computation could and could not reach, wrote that the heart has its reasons which reason does not know (Pascal, 1670). Robert Bellah and his colleagues took Tocqueville’s phrase for the American case, habits of the heart, to name the moral formation a people hold in common and can lose without noticing (Bellah et al., 1985).

Yantra: the mechanism through which intelligence operates, from cosmic systems to technological devices. Krishna’s revelation dismantles the false opposition between spirit and technology. Beings have always been yantrārūḍhāni, seated upon cosmic machinery. This sense of mechanism runs unbroken from one age to the next; the cart’s wheel, the printing press, the steam engine, and now artificial intelligence. Each one is a yantra, a mechanism through which intelligence works. The world itself, as a series of linkages and causation, is a machine, where the parts operate under the influence and constraints placed by adjacency. The question is never whether to engage a machine, but whether divine intelligence guides that engagement through hearts prepared for discernment.

Technology itself, of course, is not the problem. Yantra has always been part of how the divine operates. What has gone wrong is that we have built ever more powerful yantra while starving the cultivation of hṛdaya, machines of great capability with emptied hearts behind them.

AI systems are the most powerful yantra ever built, and they shape human behavior at a scale no earlier one approached. But who designs them? Engineers, mostly, trained where ethics is an elective if it is offered at all. They work for entrepreneurs whose schools teach that value means shareholder return, inside policy regimes advised by consultants with no room in their frameworks for transcendent purpose.

Conceding the Concentration

The most powerful thinking machines ever built will concentrate power. There is not much disagreement on this statement. The labs building frontier AI say so as readily as their critics do. The disagreement is only about whether to welcome it, and what moral justifications are offered while taking a position.

The concentration has a physical form. A frontier model is trained on a quantity of specialized chips, housed in datacenters that draw the power of a small city, that perhaps a dozen organizations on earth can assemble and afford. The capital behind a frontier training run has grown at roughly two and a half times a year since 2016, from tens of millions to the hundreds of millions, and crossed half a billion dollars in 2025 on its way to the billions (Cottier et al., 2024; Epoch AI, 2024). The firms that can raise that money are largely the firms that already hold it. The data was scraped once by a few labs, at planetary scale, and cannot be scraped again because the world has noticed. The few thousand researchers who can build at this level move among that same handful of labs and nowhere else. Compute, capital, data, and talent have each concentrated in the same places, a frontier held by a few.

These facts settle the whether and the where, but not the how. How is a concentration of power this large controlled, how is it constrained, and how is it put to work for the welfare of the society it is reshaping?

An Old Question

Our machines are new, so we assume that we might be the first ones required to imagine and ask this question. But constraining a concentration of power is the oldest task a civilization has, the second being to bend it toward the common good.

Every civilization has met a concentration of power and had to decide what to do about it. The form changes from age to age. It has been a king, a priesthood, an empire, a merchant house, a bank, a bloc of capital. How is a power this large kept honest, and how is it bent toward the welfare of the many rather than the comfort of the few who hold it? Civilizations that answered it well lasted. The ones that answered badly became the ground the others built on.

So, in this telling, AI is merely the newest holder of concentrated power, and the question it raises is the one human societies have had to work at for as long as there have been societies. If the how were genuinely new, nobody would have standing to answer it, and the future would belong by default to whoever built the machine. Because the how is old, the institutions that have worked at it longest would have something to say, or should.

All Roads Lead to Rome

The Catholic Church has built the most developed religious engagement with AI so far. Much of the work runs through Father Paolo Benanti, a Franciscan friar trained in engineering and holding a doctorate in moral theology, whose doctoral research examined cyborgs and human enhancement, who coined the term algor-ethics for the ethical study of algorithmic systems, and who advises both the Vatican and the United Nations body on AI (America Magazine, 2024). He can speak to a technologist in the technologist’s own patois while bringing to the technologist’s views a theological depth that secular ethicists cannot match.

The Rome Call for AI Ethics, first signed in February 2020 by the Pontifical Academy for Life together with Microsoft, IBM, the Food and Agriculture Organization, and the Italian government, commits its signatories to transparency, inclusion, responsibility, impartiality, reliability, and security (Pontifical Academy for Life, 2020). Cisco joined in 2024 (Cisco, 2024), and the signatory roster has since grown to more than ninety organizations (Rome Call for AI Ethics, n.d.). In July 2024, at Hiroshima, eleven world religions added their names, Buddhist and Hindu and Zoroastrian and Bahá’í among them, which put Abrahamic and Dharmic signatures on the same AI ethics document (Pontifical Academy for Life, 2024). In May 2026, Leo XIV signed his first encyclical, Magnifica Humanitas: On Safeguarding the Human Person in the Time of Artificial Intelligence, on the 135th anniversary of Rerum Novarum; it was released ten days later (Leo XIV, 2026). The document holds that technology is never neutral, because it takes on the characteristics of those who devise, finance, regulate, and use it. Anthropic’s co-founder Chris Olah was at the Vatican presentation, invited from the AI side (Fortune, 2026). This is not gesture.

Behind the diplomacy sits a university infrastructure. Notre Dame’s DELTA Network, funded by a Lilly Endowment grant of $50.8 million and shaped through more than two hundred stakeholder conversations, grounds AI ethics in named principles of dignity, embodiment, love, transcendence, and agency (University of Notre Dame, 2026). The Notre Dame-IBM Technology Ethics Lab, a ten-year partnership worth twenty million dollars, places theologians and philosophers in working contact with IBM researchers (University of Notre Dame, 2020). Santa Clara University, inside Silicon Valley, runs the Markkula Center for Applied Ethics, which joined the industry’s Partnership on AI in 2017 alongside Amazon and Google and Apple, and in 2026 an NVIDIA executive endowed a new center there for applied AI and human potential (Santa Clara University, 2017, 2026). This is an institution doing the ancient task in earnest, placing itself beside a concentrated power and holding it to a floor, a minimum beneath which it may not go.

This one tradition has moved first and built real infrastructure, to its credit. It also means that one tradition wrote the grammar in which the rest of the conversation is now expected to take place. The Rome Call’s principles read as neutral, and they are also, unavoidably, the principles of the institution that drafted them. A tradition that signs later signs language it did not shape. And the objection is not only about who held the pen. Look at what the six commitments cover. Transparency, inclusion, responsibility, impartiality, reliability, and security are all constraints on how a system may behave, and not one of them says what a system is for. The Rome Call is a brake, written by the tradition that arrived first, and a tradition holding action for the holding-together of the world as a positive duty would have pressed for the other half. It was not in the room to press. This is nobody’s crime. It is what happens when one party has the diplomatic reach and the centuries of practice, and moves while the others are still deciding whether to. The table is set by a party of one.

Most of the world’s other traditions have produced no institutional response to AI worth the name. The Hindu silence has a different cause, and a different cure.

Hindus are among the people building artificial intelligence. No region of the United States is as Indian as the Bay Area: 488,000 residents of Indian origin across the nine counties in 2024, against 18,000 in 1980, roughly six percent of the region and the highest share of any American metropolitan area, with Fremont alone at thirty percent of the city (Cheang, Ravani, & Devulapalli, 2026). Two-thirds of Silicon Valley’s technology workers are foreign-born; a quarter were born in India, the largest share after those born in the United States, and among tech workers holding a bachelor’s degree or higher, twenty-three percent came from India against seventeen percent born in California (Joint Venture Silicon Valley, 2025). Thirty percent of the region’s Indian American workers are software engineers or computer scientists. They are the second-largest group in the region’s executive ranks after white Americans, and the highest-earning community in it; Google and YouTube are both run by Indian Americans.

Not all of them are Hindu, of course. The community carries several faiths, and the same suburbs hold gurdwaras and masjids alongside the temples. But the Hindu presence in this industry is large, senior, and by now three decades old. This is not an industry Hindus failed to enter. They entered a generation ago and stayed.

What is missing is one particular institutional form, and the comparison with Rome shows precisely which. The Pope can issue encyclicals because he stands at the center of a hierarchy that authorizes him to speak in its name, and even then, he speaks for Catholics and not for all Christianity; no comparable document came from the Orthodox communions or the Protestant ones. Hinduism has no such center and no mechanism for producing such a document. Its authority is distributed across sampradāyas, maṭha lineages, and living teachers, and that distribution is a feature of the tradition rather than a defect in it. So the silence is real only where documents are signed. Hindu thinkers are working on artificial intelligence now, this paper among them.

That is a different problem from absence, and so must take a different remedy. A tradition short of people needs recruitment. A tradition whose authority is distributed needs coordination, which is harder in one respect and far easier in another: the capacity already exists, and it is already inside the industry. What it has never had is the ability to speak once and be heard as the tradition speaking. A Hindu representative did sign at Hiroshima, and the name appears nowhere in the event’s own published record. That is the rub. Someone did come. What no one could say was on whose authority.

The space the traditions left has not stayed empty. A secular movement has moved into it. The journalist Brendan Bordelon documented in 2023 how a network of adherents to Effective Altruism had spread through Washington and come to shape how the White House, Congress, and the think tanks approach the technology (Bordelon, 2023; Gebru, 2022). EA-affiliated organizations, among them Open Philanthropy, the Future of Life Institute, and the Center for AI Safety, staff a large share of the safety teams at the frontier labs and the policy desks in government, and the 2023 crisis on OpenAI’s board was widely read as a contest between this movement and commercial pressure. It is well funded, well staffed, and serious, a real response to a real danger, and it deserves to be met as one.

But it is a secular calculus that has no need for theology, a telos, or moral formation. A view that treats consciousness as an accident of matter in motion has nowhere to put formation, because on that account there is nothing there to form. The framework the AI researchers Timnit Gebru and Émile Torres have named the TESCREAL bundle (Gebru & Torres, 2024) weighs expected value across vast imagined futures, and a calculus of that kind discounts present harm against speculative future benefit. It has nowhere to put the things a wisdom tradition carries: the dignity of a person, the formation of character, the holding-together of a society. Most traditions would reject the longtermist calculus on principle. The Hindu refusal to cause present suffering on a wager about the future, the Jewish priority of immediate human welfare, the Christian preferential option for the poor, the Islamic weighing of the actual over the hypothetical good all cut against it.

This is Ex-Deus Machina. The old stage device was the deus ex machina, god from the machine, lowered by crane to resolve a plot that could not resolve itself. What runs now is god out of the machine: nobody descends at the end, and the machine itself arrives looking as though nobody built it. It sets the terms by default.

One tradition has answered in strength but alone; most have not answered, and a secular movement occupies the ground the others left open. None of that is the response the moment most needs, which is religious, strong, and coordinated at once, many traditions at one table with none setting it alone. This paper proposes that the response have a name: Deus Et Machina, God-and-machine, theology-and-technology, hands-and-heart. The god came from the machine once, lowered to end a plot that could not end itself, and has since gone out of it altogether. The name proposes God with it. The Gītā staged the third arrangement long before either: Krishna does not descend at the last act to settle Kurukshetra, he rides in the chariot from the first and forms the man who has to fight it.

Why the Traditions Have Been Standing

Religious institutions are slow. That is true without argument. They deliberate for years and argue internally for decades, and beside a technology that moves in quarters their pace looks absurd. Several of them are also, by their own admission, in poor order. Why would anyone hand them a place at this table?

And yet the slowness turns out to be part of the answer. An institution that bent to every shift in the weather would have dissolved long ago for having no center to hold to. An institution that locked every member to one fixed creed against the changing world would have been overtaken and discarded. The traditions still standing have done neither. They hold a center, and they keep adapting to the world around it, slowly, through argument.

Sanskrit has an old name for that double motion. Sanātana dharma, usually rendered the eternal way, is a compound of two ideas held at once: sanātana, the perennial, the part that holds, and dharma, the order that keeps meeting the world as the world changes. Together they name a thing that keeps a fixed center and still moves. It is the name Hinduism gives to its own tradition, and it is also a plain account of how any tradition lasts: something kept, and something continually reworked around it. Their deliberation is that reworking.

The traditions are not being recommended here as clean. They have the longest experience of concentrated power because they have stood on every side of it: they have held it, abused it, been broken by it, and reformed themselves after each break. The Roman Church was the check on the empire in the fourth century, when Ambrose refused communion to Theodosius after the Thessalonica massacre, and itself the throne by the thirteenth, when Innocent III made and unmade kings; the Reformation broke that throne, and the Council of Trent reformed what survived, slowly, over decades. The pattern recurs in every tradition that has lasted long enough to be tested. That history is the qualification, not the embarrassment it gets apologized as. An institution that has caught the disease of power and survived it knows the disease from the inside, but a three-year-old laboratory has no such immunity, humility, or sagacity.

This is the catch. The survival machinery runs on generational time, and AI does not. These traditions have always taken three or four generations to digest a civilizational change and answer it well. They no longer have three or four generations. The foundational decisions are being made now, inside a span of years. The instrument that carried them across millennia is, for the first time, too slow for the thing in front of it. The problem is the absence of time, not of wisdom.

Unasking the Questions

A well-framed question carries the seeds of its answer. If the answer does not sit well, it would help to change the question it is responding to.

A monk asked Joshu whether a dog has Buddha-nature. Joshu answered mu, neither yes nor no, refusing the question as it stood. Robert Pirsig called this unasking the question (Pirsig, 1974). A no would have accepted the question and answered it. Mu refuses the frame in which no is an answer. Several of the live debates in AI ethics need mu before they need an answer.

Universal basic income (UBI) is offered as the answer to AI taking the jobs. The question has already conceded that human participation in the workplace is over and argues only over the size of the compensation. A job, in that frame, is a wage. But a job was always more than a wage; it was where a person was formed, was needed, had a day that was the pattern of their lives. UBI can see the loss of income but not the loss of formation, the sense of self that work was building. Marienthal, a village outside Vienna, is the classic case. Its textile mill closed at the start of the Depression and some three families in four lost their living. Marie Jahoda and her colleagues went to study it and found that even with relief payments coming in, the days came apart: the men had time they could no longer use, and the place became what the researchers called a weary community (Jahoda, Lazarsfeld, & Zeisel, 1933/1971). What is missing shows in the asking: a real answer to displacement has to replace what work was beyond the wage, the meaning and the formation and the dignity of being needed. The question was only half asked, and UBI answers only that half.

The moral status of AI is argued as the question of whether the systems are conscious, whether they can suffer, whether they are owed rights (Long et al., 2024). The moral attention runs toward the machine, fast, while the standing of the humans being shaped by the machine is treated as long settled and not very interesting. That is a misallocation of moral attention, and it asks for the prior question to be set back in front: what a human person is, and what is owed to one. The traditions have precise and ancient answers to exactly that. Moral attention is all you need.

Whether to build advanced AI at all has been narrowed to a binary, build or pause, and settled by a calculation, the probability of catastrophe multiplied by its scale (Bostrom, 2014; Future of Life Institute, 2023). The binary produces a sum that weighs the fate of human civilization. Discernment is judgment that weighs what will not compute. The binary needs a third term, the guidance under which the building happens.

Algorithmic fairness is pursued as formulas that contradict each other; no system can satisfy every fairness metric at once (Kleinberg et al., 2017; Chouldechova, 2017). The case that made this concrete was COMPAS, a risk score used in American courts. In 2016 ProPublica found that among defendants in Broward County, Florida, who were never rearrested, Black defendants had been labeled higher risk at nearly twice the rate of white defendants (Angwin et al., 2016). The vendor replied that a given score carried the same chance of rearrest for both groups. Both claims held, and the mathematics shows that when two groups are rearrested at different rates, no score short of a perfect one can satisfy both. Fairness will not reduce to a metric. A value that resists formalization cannot be installed and cannot be audited into being. It can only be carried, by formed judgment inside the process. Formation is what the fairness formulas cannot encode.

What is owed to the people whose work trained the models is posed as a property dispute: who owns the inputs, who owns the outputs, what license is due. This is an ownership frame laid over things that may not be property at all, a craft, a canon, a living tradition scraped without anyone being asked. What is owed to a culture whose work was taken without its consent runs deeper than a licensing fee, into the older matter of consent and dignity, imposed on parties who had no say in it. Te Hiku Media, a Māori broadcaster in New Zealand’s far north, gathered more than three hundred hours of spoken te reo Māori, the Māori language, from its own community, turned away offers to buy the recordings, and keeps them under a license of kaitiakitanga, guardianship, which keeps the data in the service of the people it came from (Hao, 2022).

Open versus closed model release gets treated as a single dial between two goods, safety on one side and access on the other, and the argument runs over where to set it. What the dial cannot represent is that an open model in formed hands and an open model in unformed hands are not the same object. The dial is a two-good tradeoff missing its third term, the formation of the people on either side of the release.

Model ownership gets framed as a geopolitical question: AI is a strategic asset, a few nations hold the frontier, and everyone else either builds a sovereign model or lives as a customer. The question has fixed the unit as ownership and the owner as a nation. But a model that shapes how a billion people think is infrastructure rather than an oilfield, and infrastructure carries the values of whoever built it as though they belonged to no one in particular. When DeepSeek released its R1 model in January 2025, users around the world found that its chatbot declined to answer questions about what happened in Tiananmen Square in June 1989 (CNN, 2025). A model owned by one country delivers that country’s anthropology and values to every other country as the default setting, and that does not stop being colonial when the owner is friendly, or when the owner is your own nation exporting to its neighbors. The question pushes past ownership: the real question is what the model answers to, and whether that floor was set by more than one civilization.

Set the implied answers side by side. UBI was short of meaning and formation. The rights debate had mislaid the dignity of the person. The build-or-pause calculation had no room for discernment. Fairness asked for formation again. The creators’ question asked for consent and dignity, and the open-or-closed dial was missing the formation of the people it hands the model to. The ownership question asked for a floor that no single civilization sets alone. Seven debates, in seven different literatures, argued by people who mostly do not read each other, and they surrender the same short inventory: formed people, the dignity of the person, discernment, consent, a floor held in common. Each one is something a wisdom tradition has cultivated for a very long time, and a secular calculus structurally cannot supply.

Seven different questions that must be unasked before they can be answered.

The Wheel and the Yoke

The answer those questions converge on has the shape of a wheel. A wheel is the oldest machine that is still at the center and fast at the rim at the same time, the only way to move fast and not break things. Deus Et Machina is built around that wheel.

Seen whole and in one turn, it works like this. A dozen traditions put their names to three sentences about what a human being is and what may not be done to one. Each sends people it has already formed, over years and not months, into the labs and the agencies where models are cleared for release, on the staff and in the meetings. Those people carry the three sentences into decisions about particular systems in the quarter those systems ship, and they do not return to the coalition for permission each time. When a company crosses the line it signed, they say so publicly and by name. And where a system could serve the sick, the dying, or the unschooled, the traditions do not have to ask a lab to build it. They already run the hospitals, the hospices, the schools, and the shelters. They can commission the work, pay for it, and deploy it in their own institutions, which makes them a customer with standing rather than a petitioner. These institutions are already large. Every day, more than one in seven patients in the United States is cared for in a Catholic hospital (Catholic Health Association of the United States, n.d.), and the Akshaya Patra Foundation, begun by the International Society for Krishna Consciousness (ISKCON) in Bengaluru, serves a hot midday meal to more than two million schoolchildren in India every school day (Akshaya Patra Foundation, n.d.). None of that waits on the traditions agreeing about God.

A center must be changeless. Changelessness cannot be built, because it is not a feature anyone constructs, at any speed or budget; it comes from timeless age alone. A center is changeless because it has already held for a long time. So, the wheel must be made in two ways at once, by inheriting the hub and constructing the rim.

Borrow the hub from the traditions, each of which is already old. From what they already agree on, make a short charter of what AI development must answer to. To wit: human beings are not raw material; a person is not an information process to be optimized; no machine should operate beyond the reach of that claim. The charter would not be a creed that forces the traditions to converge, which would be domination under another name. It is almost nothing, the least that everyone could agree on. Laozi found the same principle in the same object: thirty spokes share one hub, and it is the emptiness at the center that makes the wheel useful (Laozi, n.d., ch. 11). A charter that said more would bind less, and hold fewer. The hub is thin, not hollow. What little it carries it carries absolutely.

The hub is also a yoke, restraining and aiming at once. Restraint the coalition can start on now: a company that signed the charter can be named against a specific model, in the quarter it ships, by people who have read the system card. Aiming is harder and almost nobody attempts it. It means naming what these systems should be built for and putting money and people behind those uses: care for the sick and the dying, teaching, the languages and texts about to be lost, work that holds a community together rather than work that holds its attention. Rājadharma binds the ruler under the law instead of above it. Loka-saṃgraha, from the same Gītā (3.20), is action for the holding-together of the world. The first without the second is a brake.

The spokes are the traditions, and every spoke runs the same distance whatever the size of the tradition behind it. A hundred thousand adherents count at the hub for what a billion count, and the Parsis, whose signature was already on the Rome Call at Hiroshima, sit where Rome sits. That is a real cost to the large traditions and it is the only reason the small ones would come. A spoke is a standing delegation with a name, a convener, and a budget line, assembled from bodies the tradition already runs, and it can be constituted in a season by people who already know each other. The delegations never merge. Nobody converts, no creed is forced, and each one answers for its own tradition and no other. Whether the multi-tradition claim is a hope or a working fact is settled by how many delegations are actually named.

Only the rim gets built: small funded teams embedded in the AI labs and the standards bodies, taking specific positions on specific models, moving in quarters. They do not return to the whole coalition for permission before each move. The coalition agreed to the hub once, and the rim acts from it. That delegation lets the wheel run fast without breaking the agreement that holds it together. The scale is small on purpose. Two or three people inside each of the dozen organizations that decide anything, thirty in all, or one hundred twenty if the thinking is to be properly saturnine and slow, a counterweight to the fast thinking everywhere else in the building. Śani keeps the accounts, and wears a rim.

They would be paid out of budgets the traditions already run and the foundations already giving to their work, or funded out of the charitable giving arms of the technology companies themselves, on the terms Notre Dame took IBM’s money and Santa Clara took NVIDIA’s: endowment rather than annual grant, and no say over what gets said. An entire operating layer for less than the cost of a single training run is a bargain in all languages.

The rim still has to earn its place inside the labs, and trust there runs in a second direction. The technology companies are not waiting for the traditions to organize; they ship models in quarters, and they will not pause while a coalition forms. But the practitioners now staffing their trust, safety, and governance roles report a shortage they cannot fill from the usual pipelines: people with the judgment to make calls that no rule settles in advance. That is the opening. A tradition enters a lab by offering formed judgment as working input rather than commandments or public relations. Trust with the companies is built slowly and by showing up. What the traditions carry that a lab cannot hire is the practice of slowness: silence, reflection, and timeless inquiry.

Keeping the wheel turning takes three older words, and each names a requirement rather than a virtue.

Location is where the people sit. Faculty and fellows on the staff roster inside the AI labs and the agencies that regulate them, present in the meetings where a model is cleared for release, not commenting from outside once it has shipped.

Vocation is why they are there, put plainly enough that an engineer can repeat it without irony. Their work is to bring judgment the room lacks to decisions the room is already making, which is a different job from auditing afterward and a different job from blessing.

Formation is how such a person is made, and it is the part nobody can shortcut. A tradition forms people the way it always has: a teacher, years of practice, a community that corrects, texts read slowly and returned to. There is no accelerated version and no certificate. This is why a tradition can supply what a training course cannot, and also why the supply is small, slow, and has to be started before it is needed. A person formed this way is a spoke under tension, bound at once to an ancient center and a fast edge, and formation is what keeps that spoke from going slack.

The hospital chaplain is the working precedent for all three. American hospitals already employ people formed by religious traditions, on their own payroll, in secular and religious hospitals alike. A board-certified chaplain brings seventy-two graduate hours of theological study and four units, sixteen hundred hours, of supervised clinical pastoral education (Board of Chaplaincy Certification Inc., 2025) and is called to the bedside when a family must decide what no protocol decides for them. The chaplain does not run the hospital, and hospitals keep paying for one. The rim proposed here is close kin to that post, placed where the models are made, with the added standing to name a company that breaks the charter.

What such a person carries into the lab is hṛdaya. This paper opened on the complaint that we have built ever more powerful yantra while starving the cultivation of the heart that should guide it, and there is no repair for that except putting formed people where the yantra is made. They are not there to slow the work, and they are not there to mourn it. They are there so that a human being is in the room while it is being decided what a human being is.

Repurpose, Do Not Build From Scratch

The wheel needs a hub that is already old, and that rules out building one. An institution with standing is an institution with age, and age is the one thing a founding cannot supply. So the hub is found rather than founded.

The constraint is practical and doctrinal. Creation out of nothing is a Western premise. The Sāṅkhya account is satkāryavāda, where the effect is already present in its cause, and pariṇāma, the unfolding of what was there rather than the arrival of what was not. Nothing on this view begins; things turn into what they were carrying. An institution follows the same rule. It is never founded out of nothing either, only inherited and pointed somewhere new.

The wheel’s hub needs a standing multi-tradition convening, the kind of body whose explicit purpose is to gather the world’s religious traditions in periodic conversation. For example, the Parliament of the World’s Religions, founded in 1893 and revived in 1993, is the longest-running such body, and the 1893 Parliament in Chicago is where Vivekananda walked onto the world stage and Hinduism announced itself to the interfaith world (Parliament of the World’s Religions, n.d.). That history is also why the missing Hindu statement reads as a matter of form rather than of will. The Parliament is one example of the category, not the only one. Some convening of this kind needs to be pointed at the problem, and which one is a second question.

Between convenings, the wheel needs something operational, because a body that meets every few years cannot keep pace with a technology that ships in quarters. A multi-tradition academic node can take that role: a place where scholars and practitioners of several traditions are already housed together, and where the rim work, embedding fellows in labs and regulatory bodies, can be staffed and supervised. Here one could point to the Graduate Theological Union in Berkeley, which holds Christian, Jewish, Islamic, Buddhist, and Hindu studies together within reach of both Silicon Valley and a major AI research center. Others exist, and any of them would serve.

The spokes need each tradition’s own institutional capacity, already organized enough to bear load. The Hindu spoke, unassembled rather than absent, is a useful illustration of how a spoke gets assembled from pieces that already exist. The Hindu American Foundation has policy capacity and access; the Ramakrishna Mission has spiritual authority and a century of institutional discipline; ISKCON has organizational reach across continents; the Shankaracharya peethams hold traditional authority of the oldest kind; the Indian institutes of technology and management have the technical depth to read what the labs are actually doing; the Hindu Spiritual Care Institute has board-level capacity to convene. None of these is, by itself, the Hindu answer to AI. Together, coordinated, they are the spoke. The work is not to create Hindu institutional capacity from nothing; it is to point capacity that already exists at a problem it has not yet been turned toward. Every other tradition has its own such inventory, and the same work to do with it.

This should not be easy. Welding a periodic convening, an academic node, and a set of traditional institutions into something that operates at quarterly speed runs into real coordination costs, political friction within each tradition, and the routine difficulty of moving any old institution toward a problem it was not designed for. The bodies that exist are slow, sometimes territorial, often divided internally, and rarely set up to act in concert. None of the bodies named above are in a coordinating relationship with each other today, and constructing those relationships is itself the work of years before the wheel turns once. None of that is an argument for founding something new, which would face every one of those costs without the inheritance. It is an argument for taking the friction seriously and budgeting for the years that working through it will take.

Facing the Objections

We expect that a proposal to seat the wisdom traditions beside the governance of AI will invite hard objections, and three of them deserve a direct answer.

The most familiar objection is the historical record. Interfaith coordination has been tried many times, and its characteristic product is a watered-down statement that everyone can sign because it commits no one to anything. Councils of religions tend to dilute toward the lowest common denominator, and a coalition assembled to govern AI would, by this reasoning, produce a bland charter and no action. The objection has force, and the wheel is built against it. What dilutes an interfaith body is the attempt to agree on doctrine, on the nature of God or the soul or the good. The wheel never asks for that. The hub carries a charter and not a creed, the spokes are meant to stay separate, and the only thing every tradition signs is those three clauses, whose whole content is that a person is not raw material, a resource. An agreement that thin does not dilute, because there is nothing beneath it to dilute toward. And the action does not wait on consensus, because the rim is delegated. A coalition dies as a committee when it makes the many decide the fast things, and this design never does.

Weightier still is the objection that the proposal is religious overreach, that the governance of a secular technology should stay secular, and that inviting the traditions in is a step toward theocracy. But what the traditions are asked to do is stand beside the power and hold it to society’s benefit, which is the work of a check rather than a sovereign; rājadharma is precisely the name for a power held to account rather than a power exercised. The deeper trouble with the objection is its picture of the alternative. The alternative to religious involvement is the secular calculus already installed, the longtermist expected-value framework, which is itself a contested account of value rather than a neutral one, now imposed on entire civilizations that had no input into it. The real choice lies between one value-framework and several, and a floor held by many traditions is wider and less particular than a calculus held by a single movement.

The last objection is that the traditions disagree too deeply for any shared floor to be real, that Hindu and Muslim and Christian and Buddhist metaphysics cannot be reconciled, and that pretending otherwise is a fiction. But the argument does not rely on a convergence of cosmologies; it is the convergence the ethics questions already showed, seven debates arriving independently at the same short inventory. A tradition does not have to share another’s account of God to share the claim that a human being is not an information process to be optimized. The spokes of the wheel stay distinct exactly so that the traditions never have to agree on what they cannot agree on. They are asked to agree on the charter, and the charter is small enough, and human enough, that they already do. Humans should decide what it means to be human.

These objections are real; each can be argued for a decade. We do not have decades to decide.

Implications for Developing Nations

The argument has a particular weight for developing nations, because the danger reaches them in a specific form. The dominant model of AI adoption is the import of Silicon Valley systems together with the assumptions built into them. A kind of digital colonialism (Kwet, 2019; Couldry & Mejias, 2019; Birhane, 2020): one civilization’s account of human flourishing arriving as infrastructure in places whose own traditions hold a different account.

The two accounts are not small variations on each other. Silicon Valley’s implicit anthropology treats a human being as an information processor whose wellbeing can be read off engagement metrics and improved by algorithmic intervention. The dharmic anthropology treats a human being as a conscious presence whose flourishing turns on relationship, meaning, purpose, and discernment, none of which a metric can hold. A model trained inside the first anthropology and deployed inside a community that lives by the second does not arrive as a neutral tool but as a replacement of the entire system of being.

A developing nation has two poor options and one good one. It can refuse the technology and fall behind, which competitive pressure makes untenable. It can import the technology wholesale and let its embedded anthropology do its slow work, which is the colonial outcome we are already familiar with. Or it can adopt the technology while building the institutional capacity to govern it from its own wisdom; the only path that keeps the adoption from becoming a surrender. India is the test case. It is the world’s most populous democracy, a fast-rising center of AI development, and the home of the dharmic traditions, and it currently tends to import Silicon Valley frameworks rather than draw on the resources under its own feet. The wheel proposed here is, among other things, a way for nations in that position to shape what AI answers to rather than only receive what it delivers.

Conclusion

Oppenheimer’s failure was not a failure of wisdom. He had the wisdom. It was a failure of infrastructure: one formed conscience and no institution to carry it. The decisions being made now about AI are larger than the ones he faced, and they are being made, again, without an institution able to hold wisdom against power.

The wheel of response can be built. Most of it does not need building from scratch; its hub is inherited from traditions older than any country, its spokes are those traditions themselves, and only the rim must be made new, well-formed people put into well-funded companies. What remains is the decision to turn it. Power can be left to concentrate unyoked, and history is unambiguous about how that ends every time. Or it can be yoked to a charter that many civilizations hold, and made to pull the cart toward the holding-together of the world.

Our window for making the choice is narrow. The next two or three years will settle who sits on the safety and policy boards inside the labs, what charter the companies sign and treat as binding, and which voices the governments call when they decide. Those decisions will keep AI’s timeline, not the traditions’.

Machines made from human capability have outrun the wisdom available to govern them. The institutions that could carry that wisdom already stand, older than the machines and older than most of the countries. What remains is to point them at the heart of the machine.

References

Akshaya Patra Foundation. (n.d.). About us. https://www.akshayapatra.org/about-us

Altman, S. (2024). The age of intelligence. OpenAI.

America Magazine. (2024, January 19). Meet the Franciscan friar advising Pope Francis on AI ethics. https://www.americamagazine.org/faith/2024/01/19/friar-paolo-benanti-artificial-intelligence-ai-ethics-vatican-246989

Angwin, J., Larson, J., Mattu, S., & Kirchner, L. (2016, May 23). Machine bias. ProPublica. https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing

Atomic Heritage Foundation. (n.d.). Trinity Test, 1945. https://ahf.nuclearmuseum.org/ahf/history/trinity-test-1945/

Bellah, R. N., Madsen, R., Sullivan, W. M., Swidler, A., & Tipton, S. M. (1985). Habits of the heart: Individualism and commitment in American life. University of California Press.

Birhane, A. (2020). Algorithmic colonization of Africa. SCRIPTed, 17(2), 389–409. https://doi.org/10.2966/scrip.170220.389

Board of Chaplaincy Certification Inc. (2025). BCCI certification manual. Association of Professional Chaplains. https://www.apchaplains.org/bcci-site/about-bcci/bcci-certification-manual/

Bordelon, B. (2023, October 13). How a billionaire-backed network of AI advisers took over Washington. Politico. https://www.politico.com/news/2023/10/13/ai-influence-effective-altruism-open-philanthropy-00121362

Bostrom, N. (2014). Superintelligence: Paths, dangers, strategies. Oxford University Press.

Catholic Health Association of the United States. (n.d.). Facts and statistics. https://www.chausa.org/newsroom/facts-and-statistics

Cheang, K. L., Ravani, S., & Devulapalli, S. (2026, May 26). Their meteoric rise reshaped the Bay Area and powered Silicon Valley. Is it at an end? San Francisco Chronicle. https://www.sfchronicle.com/projects/2026/indian-american-sf-bay-area/

Chouldechova, A. (2017). Fair prediction with disparate impact: A study of bias in recidivism prediction instruments. Big Data, 5(2), 153–163. https://doi.org/10.1089/big.2016.0047

Cisco. (2024, April 24). Cisco joins Rome Call as AI ethics debate widens. Cisco Newsroom.

CNN. (2025, January 29). DeepSeek is giving the world a window into Chinese censorship and information control.

Corbin, H. (1969). Creative imagination in the Sufism of Ibn 'Arabī (R. Manheim, Trans.). Princeton University Press.

Cottier, B., Rahman, R., Fattorini, L., Maslej, N., & Owen, D. (2024). The rising costs of training frontier AI models. arXiv. https://arxiv.org/abs/2405.21015

Couldry, N., & Mejias, U. A. (2019). Data colonialism: Rethinking big data's relation to the contemporary subject. Television & New Media, 20(4), 336–349. https://doi.org/10.1177/1527476418796632

Epoch AI. (2024). How much does it cost to train frontier AI models? https://epoch.ai/blog/how-much-does-it-cost-to-train-frontier-ai-models

Fortune. (2026, June 3). Who is Chris Olah? The Anthropic cofounder the Pope chose to sit beside him at the Vatican. https://fortune.com/2026/06/03/who-chris-olah-anthropic-cofounder-atheist-pope-leo-vatican-ai/

Future of Life Institute. (2023, March 22). Pause giant AI experiments: An open letter. https://futureoflife.org/open-letter/pause-giant-ai-experiments/

Gebru, T. (2022, November 30). Effective altruism is pushing a dangerous brand of "AI safety." Wired.

Gebru, T., & Torres, É. P. (2024). The TESCREAL bundle: Eugenics and the promise of utopia through artificial general intelligence. First Monday, 29(4). https://doi.org/10.5210/fm.v29i4.13636

Hao, K. (2022, April 22). A new vision of artificial intelligence for the people. MIT Technology Review. https://www.technologyreview.com/2022/04/22/1050394/artificial-intelligence-for-the-people/

Jahoda, M., Lazarsfeld, P. F., & Zeisel, H. (1971). Marienthal: The sociography of an unemployed community. Aldine-Atherton. (Original work published 1933)

Joint Venture Silicon Valley. (2025). 2025 Silicon Valley Index. https://jointventure.org/silicon-valley-index

Kleinberg, J., Mullainathan, S., & Raghavan, M. (2017). Inherent trade-offs in the fair determination of risk scores. In 8th Innovations in Theoretical Computer Science Conference (ITCS 2017) (pp. 43:1–43:23). https://doi.org/10.4230/LIPIcs.ITCS.2017.43

Kwet, M. (2019). Digital colonialism: US empire and the new imperialism in the Global South. Race & Class, 60(4), 3–26. https://doi.org/10.1177/0306396818823172

Laozi. (n.d.). Tao te ching (D. C. Lau, Trans., 1963). Penguin Classics.

Leo XIV. (2026). Magnifica humanitas: On safeguarding the human person in the time of artificial intelligence [Encyclical letter]. Vatican. https://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html

Long, R., Sebo, J., Butlin, P., Finlinson, K., Fish, K., Harding, J., Pfau, J., Sebastián, T., Schwitzgebel, E., & Chalmers, D. (2024). Taking AI welfare seriously. arXiv. https://arxiv.org/abs/2411.00986

Muller-Ortega, P. E. (1989). The triadic heart of Śiva: Kaula Tantricism of Abhinavagupta in the non-dual Shaivism of Kashmir. State University of New York Press.

Parliament of the World's Religions. (n.d.). 1893 Chicago. https://parliamentofreligions.org/parliament/1893-chicago/

Pascal, B. (1670). Pensées, Lafuma §423 (A. J. Krailsheimer, Trans., 1995). Penguin Classics.

Pirsig, R. M. (1974). Zen and the art of motorcycle maintenance: An inquiry into values. William Morrow.

Pontifical Academy for Life. (2020). Rome call for AI ethics. Vatican.

Pontifical Academy for Life. (2024). AI ethics for peace: World religions commit to the Rome Call. Vatican.

Rome Call for AI Ethics. (n.d.). Organisations [Signatory list]. Retrieved from https://www.romecall.org/organisations/

Santa Clara University. (2017, November). Markkula Center joins international AI partnership.

Santa Clara University. (2026, April 14). Santa Clara University establishes the Cunningham Shoquist Center for Applied AI and Human Potential.

University of Notre Dame. (2020). Notre Dame, IBM launch Tech Ethics Lab. Notre Dame News.

University of Notre Dame. (2026, January). Notre Dame receives $50.8 million grant from Lilly Endowment for the DELTA Network. Notre Dame News.

Upaniṣads. (n.d.). Upaniṣads (P. Olivelle, Trans., 1996). Oxford University Press.

Vyasa. (n.d.). Bhagavad-gītā as it is (A. C. Bhaktivedanta Swami Prabhupada, Trans., 2nd ed., 1972). Bhaktivedanta Book Trust.

Ware, K. (1979). The Orthodox way. St Vladimir's Seminary Press.