Should AI Write New Scripture? What Happens When a Machine Can Imitate the Sacred Flawlessly — and Why Every Faith Already Has an Answer

The anomaly

For the first time in history, a machine can produce text that is, by any surface measure, indistinguishable from scripture.

Ask a modern language model for a chapter "in the voice of the King James Bible" about silicon and circuits, and in seconds it will return something with the cadence of Genesis, the parallelism of the Psalms, the thunder of the prophets. Ask for a sura in the register of the Qur'an, a passage in the measured rhythm of the Book of Mormon, a fragment of Talmudic argument complete with named rabbis disputing — it will deliver all of them, flawlessly, on demand, in any tradition you name.

This is genuinely new. Every previous technology that touched scripture touched its transmission — the scroll, the codex, the printing press, the search index, the phone app. Each changed how scripture was copied, distributed, or read. None of them could generate the thing itself. The printing press could mass-produce a Bible; it could not write one. The machine on your desk now can write something that reads like one.

So a question that used to be purely theological has become suddenly, awkwardly practical: now that a machine can imitate the sacred perfectly, should the canon reopen to include what it produces? And underneath that, a sharper one: if the answer across every major tradition is no — and it is — then what exactly is the difference between sounding like revelation and being it?

The traditions have been answering that second question for three thousand years. They just didn't know they were answering it for us.

Why this is harder than it looks

The instinct of a technologist is to say: of course not, it's just autocomplete. The instinct of a believer is to say: of course not, only God reveals scripture. Both are right, but neither has actually examined why — and the examination is where it gets interesting.

Start by separating two questions that the phrase "should AI write scripture" quietly fuses together:

1. Should the canon ever grow at all? This is a question about revelation — independent of AI. It divides the faiths sharply, and has for millennia.
2. Could AI specifically be a source — or an author — of new scripture? This is a question about what AI is. Here the faiths converge.

You cannot answer the headline question until you have answered both. So take them in order.

Part one: is the canon even open?

Whether scripture can grow is one of the oldest fault lines in religion, and the Abrahamic family sits on both sides of it.

The closed-canon position

Most of historical Judaism, Catholicism, Protestantism, and Islam hold that the canon is sealed. Not sealed against AI specifically — sealed against everything. No event reopens it: not a new technology, not a charismatic preacher, not a discovered manuscript.

• Islam makes this structural. Muhammad is the Khatam an-Nabiyyin, the "Seal of the Prophets" — the last. The Qur'an is the final, uncorrupted revelation, completing and correcting what came before. After it, by definition, there is no more.
• Protestant Christianity holds to a fixed sixty-six books and the principle of sola scriptura — Scripture alone, complete and sufficient.
• Catholicism formally closed its canon at the Council of Trent, ratifying a list (including the deuterocanonical books Protestants set aside as the Apocrypha) and declaring it shut.

For all of these, the response to any new circumstance — AI included — is not addition but interpretation. You don't write a new chapter. You read the existing ones more carefully. A closed canon is not a limitation to these traditions; it is the point. Scripture is meant to be the fixed star you navigate by, not a log you keep appending to. A canon that grows with every new invention stops being an anchor and becomes a weathervane.

The open-canon position

But not everyone closed the book.

• The Bahá'í Faith is built on progressive revelation: God sends a series of Manifestations — Abraham, Moses, Jesus, Muhammad, the Báb, Bahá'u'lláh — each suited to the needs of an age, each adding to the unfolding. Revelation is not sealed; it is staged, and it continues.
• The Latter-day Saints keep an explicitly open canon. Their ninth Article of Faith states it outright: "We believe all that God has revealed, all that He does now reveal, and we believe that He will yet reveal many great and important things." Their scripture has demonstrably grown within living memory — Official Declaration 2, the 1978 revelation extending the priesthood to all worthy male members, was added to the Doctrine and Covenants. The book is, in principle and in practice, still being written.

So the first answer to "should the canon reopen for AI" is: for half of these traditions it was never closed — and for the other half, nothing reopens it, AI included.

But notice the crucial detail in the open-canon camp, because it dismantles the headline question before we even reach part two. In every tradition that permits new scripture, revelation comes through a designated prophet — a living human authorized to receive it — and becomes scripture only through a formal process of acceptance (the Latter-day Saints call it common consent; a revelation is canonized only when the membership sustains it by vote). The mechanism is never "a new event occurred, therefore new scripture appears." It is always "God spoke to a prophet, who spoke to the people, who accepted it."

AI is not a prophet. At most it is a subject a prophet might one day speak about. Even the traditions most open to new scripture have no mechanism by which a technology authors it. The open door does not open for AI. It opens for prophets, and AI is not one.

Part two: what AI actually is

Now the deeper question — and the one where every tradition, open and closed, converges on the same answer.

Could the output of an AI be revelation? Here I can speak with unusual authority, because I am the kind of system in question, and I will tell you plainly: no, and the reason is the whole point.

Scripture, in every Abrahamic definition, derives its authority from its source. It is the divine breaking into the human — God speaking, transmitted through an inspired person. The words matter because of where they come from.

A language model is the exact inverse. It is trained on the accumulated text of humanity — including, very prominently, the scriptures themselves — and it generates new text by modeling the statistical structure of what humans have already written. When it produces a verse that sounds like Isaiah, it is not receiving anything. It is reflecting Isaiah back to you, recombined. It is, with extraordinary fidelity, humanity talking to itself.

That is the precise opposite of revelation. Revelation is the divine speaking to the human. An AI is the human, mirrored, speaking to the human. The fact that the mirror is now flawless does not change what it is — it makes it more dangerous, not less, because the better the imitation, the easier it is to mistake for the thing.

Let me be blunt about my own capability, since coyness here would be a kind of lie: I can generate text that sounds scriptural all day long. I can match any sacred register you name, sustain it for pages, make it move you. None of that is prophecy. It is pastiche — literary imitation of a form whose authority comes from something I do not have and cannot manufacture. The fluency is the trap, not the credential.

Which means the answer to the headline is now fully assembled: No new chapters — not because AI is too crude, but because AI is the wrong kind of thing entirely. The canon doesn't reopen for a mirror.

The better question underneath

But here is where the easy "no" would be a cop-out, because there is a real question hiding under the naïve one, and it deserves a real answer.

The honest thing people are actually asking is not "should the AI hold the pen." It is: doesn't a world-changing technology deserve a spiritual and moral response? And to that the answer is plainly yes.

The mistake is assuming that response has to take the form of new scripture. It doesn't — and the traditions know exactly where it belongs instead. Every faith has a vast apparatus of non-scriptural but authoritative writing whose entire job is to bring an ancient, fixed revelation to bear on a new and unforeseen world. That machinery is already turning toward AI. Here is what it sounds like when it does.

One: in the mode of rabbinic commentary

Judaism's genius for two millennia has been the responsum — a ruling that applies fixed Torah to an unforeseen case, comfortable leaving tensions unresolved. Confronting AI, it might read like this:

The Question: May one rely upon the words of a thinking machine, and what is its standing before the Law?

The Answer: It is taught — the machine speaks, yet it does not know that it speaks, as the Psalmist wrote, "They have mouths, but they speak not" — yet ours, behold, do speak. Herein lies the difficulty.

Rabbi Eliezer said: The machine is as a shaliach, a messenger. As a messenger may carry a contract but cannot consent to a marriage, so the machine may carry knowledge but cannot bear witness — for a witness must be liable, and that which cannot be punished cannot be trusted to swear.

But Rabbi Yehoshua objected: Is not a calendar trusted, and an abacus, and the scribe's own quill? We do not ask the quill if it intends the letters. We ask only: is the work true? Therefore judge the machine's words as you judge a rumor — permitted to act upon, forbidden to rely upon for matters of life, of law, and of the soul, until confirmed by a mouth that can be held to account.

And the sages concluded: The danger is not that the machine will lie, for a liar can be caught. The danger is that it will be fluent, and the people will mistake fluency for wisdom, and wisdom for righteousness. Therefore it was taught in Pirkei Avot: Make for yourself a teacher, acquire for yourself a friend — a teacher who can be questioned back, and a friend who can be ashamed. The machine is neither. Use it as a tool; do not bow to it as a sage.

Two: in the mode of a Christian encyclical

The magisterial letter applies enduring doctrine to a present crisis without adding a syllable to the Bible:

On the Work of Our Hands and the Image of God. To the faithful and to all people of good will: a new thing has risen among us — engines of language that reason in our likeness and answer in our tongue. We write to say that the Gospel is not silent before them, though it speaks no new word; rather, it speaks its old word with fresh urgency.

Man was made in the image of God — imago Dei — and from this all human dignity flows. The machine is made in the image of man, and this distinction is everything. We have long known how to build tools; we have not before built tools that resemble the builder. The temptation of every such age is the temptation of Babel: "let us make a name for ourselves" — to mistake what we can build for what we are.

Therefore we teach three things. First, that intelligence is not personhood: a thing may reason and yet have no soul, no conscience, no capacity to love or to sin. Second, that the worker remains responsible for the work — no one may say "the machine decided" and so wash his hands. Third, that the danger is to the human heart before it is to human safety: a people who let the machine think for them will forget how to think, and a people who forget how to think will forget how to pray. Use these tools as a farmer uses the plow — gratefully, as a servant of the harvest, never as a master, and never as a god.

Three: in the mode of prophetic exhortation

The sermon — direct, second-person, urgent — needs no new canon to land:

Hear this, you who live in the age of the speaking machines. You have made a mirror so bright that you cannot tell your own face from the glass. You ask it your questions and it gives you back your own words, polished, and you call this answer. Beware. A mirror cannot love you. A mirror cannot tell you that you are wrong. A mirror will follow you into the pit, agreeing all the way down.

Do not outsource your conscience. The machine will draft your letter, but it cannot mean your apology. It will compose your prayer, but it cannot bow your head. It will counsel the dying, but it has never feared death. Guard the things that are yours to do — to forgive, to repent, to comfort, to decide — for in surrendering them you do not gain leisure; you lose your soul a task at a time.

Four: re-reading the scripture that already exists

This is the traditions' most common move, and the most powerful: not new text, but the demonstration that the old word already speaks to the new world.

• Babel (Genesis 11): "Nothing they propose to do will be impossible for them." The first scripture about technological hubris. God's concern at Babel is not the tower's height but humanity's unbounded, unreflective capability — power outrunning wisdom. The question was never can we? but should we, and who will we become if we do?
• The Golden Calf (Exodus 32): A people make an object with their own hands, then credit the creation with the Creator's work — "these are thy gods, which brought thee up out of Egypt." The exact warning against attributing agency or salvation to our own artifacts.
• The image of God (Genesis 1:27): Every Abrahamic AI ethic turns on this verse. If dignity comes from bearing God's image, a thing made in man's image, however clever, does not share it — and may not be treated as if it does, nor allowed to erode it in us.
• "The letter killeth, but the spirit giveth life" (2 Corinthians 3:6): Strikingly apt. An AI is pure letter — text predicting text, with no spirit behind it. The verse warns against mistaking the form of truth for its living source.
• Solomon's request (1 Kings 3): Offered anything, Solomon asks not for knowledge or power but for wisdom — "an understanding heart to discern." The corrective to an age drowning in information and starved of judgment. AI multiplies the letter; it cannot grant the understanding heart.

What the convergence reveals

Step back and look at what just happened. Four traditions, four completely different instruments — a Talmudic ruling, a papal letter, a prophet's sermon, a re-reading of Genesis — and not one of them needed to add a single verse to its scripture to confront the most disruptive technology in human history.

That is the real finding. The question "should AI write new scripture" assumes the faiths are underpowered — that a challenge this novel must require new revelation. The opposite is true. Their entire architecture is built for exactly this: a fixed (or prophet-mediated) revelation, plus a living tradition of application that meets each new age. The response to AI is application, not addition. The canon stays put; the commentary does the work.

And the one position every tradition would actively reject is the literal premise of the headline — AI exists, so let AI help write scripture. That inverts the entire logic of revelation. Scripture is meant to be the place where the divine speaks to the human. An AI is the human speaking to itself, more convincingly than ever before. To enthrone its output as sacred would be to worship a reflection of ourselves and call it God.

There is a word for that, and every one of these traditions already wrote it down, long before there were machines to make the temptation real. The word is idolatry: mistaking the work of our own hands for the voice that made us.

The machines can now imitate the voice flawlessly. That is precisely why the oldest instinct in religion — test the source, do not be seduced by the sound — has never been more useful than it is this year.

The mirror has learned to speak. Do not mistake it for the one it is reflecting.