Who Owns the Worldview Inside an AI Model?

A faith-technology company has done something the IP world should notice. Gloo — a Boulder, Colorado, firm that serves churches, ministries, and Christian universities, and now trades on the Nasdaq — built a benchmark it calls Flourishing AI Christian, or FAI-C. The finding is blunt: Today’s leading large language models, tested on questions of meaning, character, and faith, come up short. On a 100-point flourishing scale, the frontier models averaged 61. On the faith dimension, they scored worst of all.

Gloo’s point is that these models carry a worldview — one users never see and never agreed to. That claim is correct. But it skips a step. A model does not acquire a worldview by magic. It acquires it from the human work it was trained on: the books, the scholarship, the sermons, the code, the arguments. If a model “thinks” a certain way, it is because it has ingested the labor of particular people. Before we ask what worldview is in the machine, we should ask who made it — and what they are owed.

Worldviews do not assemble themselves. People make the work that machines are made of.

Start with the Premise the Founders Wrote Down

I have argued for years that the right to own what one creates is not a bureaucratic nicety. It is inherent. It is a moral claim, and in America, a constitutional one. The Framers gave Congress power to secure to authors and inventors the exclusive right to their writings and discoveries. They grasped what the ancient world did not: People make new things, and the maker has a claim on what he makes. Strip the claim away and, sooner or later, you stop getting the making.

My book, To Invent Is Divine, argues the root runs deeper still — that the human capacity to create reflects the imago Dei—the image of the Creator—so ownership follows creation as a matter of justice, not merely incentive. You need not share the Judeo-Christian theology to take the practical point. The patent bargain is a 200-year-old answer to the free-rider problem. It says you may build on what others made, but not by pretending they never made it.

Which brings us back to the model.

The Training-Data Fight is an Ownership Fight

The AI-and-copyright lawsuits — the cases over whether training on protected works is infringement or fair use — get framed as fights about money. They are. But underneath, they are fights about attribution and consent, which are the two things ownership exists to protect. When a model absorbs a body of work and then produces outputs shaped by it, the people who made that work have been conscripted as uncredited, uncompensated collaborators. Their worldview is in the machine. Their names are not.

Gloo’s own discovery makes the point from the other direction. The company found that mainstream models handle Christian concepts poorly — that, asked why God permits suffering, a leading model could not give a theologically grounded answer. The likeliest reason is that the training data underweighted or flattened the relevant human work. A worldview was encoded; another was left on the floor. That is not a neutral technical artifact. It is a choice about whose labor counts.

IP law already has the vocabulary for it. We have spent decades refining what derivation means, what attribution requires, and when one person’s contribution is “significant” enough to a claimed invention to matter. Right now, the U.S. Patent and Trademark Office (USPTO) is working through the cleanest version of the question.

What the USPTO Has Said About Machines and Conception

In November 2025, the USPTO issued revised guidance on inventorship for AI-assisted inventions, rescinding its 2024 approach. The holding is clear: Only a natural person can be an inventor, and the test remains conception — a “definite and permanent idea” of the complete invention. An AI tool, however capable, is an instrument, like a microscope or a CAD program. The human who conceives it is the inventor. The machine is not.

Consider what that doctrine assumes. It assumes creative origination is a human act — that conception is the thing the law protects, and that it belongs to persons, not processes. The Office did not land there out of sentiment. It landed there because the alternative, crediting a system that recombines what persons made, would sever the link between creation and ownership that the whole system rests on.

Note the posture around it, too. Director John Squires has been emphatic that section 101 must not be used “as a blunt instrument to exclude entire technological fields” like AI, and that the doors of America’s innovation agency are wide open. The official stance is not anti-AI. It is pro-AI and pro-inventor at once. That is exactly the synthesis a faith-tech company should want — and the one my book argues the American founding already attained.

Gloo’s executive chairman is no bystander to this. Pat Gelsinger is a named inventor on eight U.S. patents in VLSI design, computer architecture, and communications, an IEEE Fellow, and a presidential appointee to the National Security Telecommunications Advisory Committee. He is also an outspoken Christian. A company run by an actual inventor, building AI it calls values-aligned, is precisely where the ownership question ought to be asked aloud — not left to litigators to settle later.

Why Gloo Should Embrace the Christian Worldview’s Human Creativity-Ownership Component

Gloo’s executive chairman is no bystander to this. Pat Gelsinger is a named inventor on eight U.S. patents in VLSI design, computer architecture, and communications, an IEEE Fellow, and a presidential appointee to the National Security Telecommunications Advisory Committee. He is also an outspoken Christian. A company run by an actual inventor, building AI it calls values-aligned, is precisely where the ownership question ought to be asked aloud — not left to litigators to settle later.

Here is the trap to avoid. A model trained to reflect a Christian worldview is still a model trained on someone’s work: theologians, pastors, scholars, writers — many living, most uncompensated and unconsulted. “Values-aligned AI” cannot mean a worldview assembled from creative labor while the ownership of that labor is treated as someone else’s problem. The same conviction that says a machine should not flatten a person’s faith says a machine should not take a person’s work. Consent, credit, and compensation are not obstacles to faith-aligned innovation. They are part of what would make it faithful.

You cannot build technology that honors the human person while treating the human creator as raw material.

Three Commitments That Would Do the Work

The faith-and-flourishing sector is unusually well positioned to lead here, by treating training data the way it already treats outputs. Three commitments would do most of the work:

Attribution by default. Build provenance into the corpus. Know whose work the model learned from and be able to say so. The technical objections shrink every year. The will to do it is the variable.

Consent as a feature, not a liability. A model that can tell a Christian publisher or a seminary its work was used with permission earns a trust no disclaimer can buy. In a sector where trust is the whole value proposition — as Gloo’s own bankers have said — that is a moat, not a cost.

Compensation that follows creation. Argue the mechanism — licensing, a collective, a negotiated rate. Do not argue the principle. If a creator’s labor makes the model more valuable, the creator has a claim on that value. That is not novel. It is the patent bargain, applied to a new kind of machine.

The Oldest Question, Asked Again

Gloo deserves credit for forcing a question the rest of the industry has been content to leave fuzzy: What worldview is in the machine, and who put it there? My amendment is only to follow it one more step. Once you concede that a worldview was authored, you are standing in the oldest territory the patent system knows — where creation and ownership are two names for human dignity.

To invent, I have argued, is divine: a share in the creative act that marks us as made in the image of the Maker. If that is true of the inventor, it is true of every creator whose work now lives, uncredited, inside AI models. The faith-and-flourishing sector, of all places, should be the first to say so.