There is a tone you have heard if you spend any time where creative people argue about AI. A respected artist declares that the people using these tools are not real artists — that their work is slop, that they should be ashamed, that what they themselves do is something sacred that cannot and must not be optimized.
I don’t think those artists are villains. I think they are craftspeople who feel something real is being lost, and the loudest tool they have to defend it is contempt.
But that tone is worth examining, because it is one side of a fight that is eating the creative world alive — and it is the wrong fight.
The loudest argument in the room
Right now, almost every conversation about AI and creativity collapses into the same two questions. Is AI-made work real art? And who owns the output?
People are pouring enormous energy into both. Artists demand that AI work be banned, labeled, or shamed. Creators who use AI tools demand that their songs and images be recognized and protected like any other work. Each side is sure the other is destroying something.
And almost everyone is arguing about copyright without noticing that copyright already moved.
Copyright didn’t disappear. It grew a condition.
Here is the part almost nobody says plainly, and the part both sides get wrong.
Start with what is genuinely settled. A purely AI-generated work — one with no meaningful human authorship — cannot be copyrighted. The U.S. Copyright Office has said so for years, and the leading challenge to that rule, Thaler v. Perlmutter, failed: the appellate court affirmed the human-authorship requirement, and in March 2026 the Supreme Court declined to hear the case, leaving that ruling standing. No court has gone the other way. If a machine generated it and no human shaped it, it has no copyright protection. Anyone can use it.
But notice the exact case that produced that rule. In Thaler, the applicant told the Copyright Office, in writing, that no human contributed to the work at all — he listed the AI as the sole author. The court ruled on that, and only that: a machine cannot be the author. It explicitly did not rule on the question that actually matters to working creators — how much human involvement is enough. That question is being litigated right now, in Allen v. Perlmutter.
And on that question, the Copyright Office has been clear, and it cuts the other way. Using AI as a tool does not strip your work of protection. Incorporating AI-generated elements into a larger work you authored does not strip the work of protection. The human-authored parts remain copyrightable. If you wrote the lyrics and arranged the song and used AI in the instrumentation, the human contribution is real — and the law still recognizes it.
So if you are an AI-assisted creator waiting to hear whether copyright abandoned you: it didn’t. What changed is the price of admission.
Copyright for AI-assisted work now comes with a condition. To register it, you have to disclose the AI-generated content and identify what a human authored. Prompts alone do not count — the Office has said that typing instructions, by itself, is not sufficient creative control. And the condition does not end at registration. Even once a work is registered, what you can actually enforce is only the human authorship: in an infringement case, the AI-generated elements are filtered out, like public-domain material, before your work is compared to the copy. So a copyright on AI-assisted work that you cannot break down — this part mine, this part the machine’s — is not just hard to obtain. It is hard to defend. The protection reaches exactly as far as the human contribution you can point to, and no further.
Ask Jason Allen how concrete that problem is. He used more than 600 iterative prompts to produce a single award-winning image, and the Copyright Office still refused to register it, finding he had not exercised enough control over the final output. He is now in federal court trying to overturn that refusal; the case is still open. Whatever the court decides, the lesson for every other creator is already clear: real human effort is not the same as provable human authorship, and the gap between the two is where protection is being lost.
That is the real problem, and it is not the one people are fighting about. The right still exists. But both getting it and keeping it now turn on the same thing — proof of human authorship — and almost no creator has any way to produce that proof, except sitting down, months after the fact, and reconstructing it from memory in a registration form.
The energy in the room is real. The frustration is justified. But it is being spent demanding that the system grant something, when the system is actually waiting for creators to document something. Copyright didn’t fail. It grew a requirement, and nobody built the tools for it.
Two walls, pointed the same way
Look at what is happening to the independent creator, and you see two pressures, not one.
The first is cultural. A respected artist tells the world that AI creators are not real, that their work is slop, that they should be ashamed. That message does real damage. It tells a person who made something they love that they are not allowed to feel good about it.
The second is structural. While that cultural message spreads, the major institutions of the creative industries are not dismissing AI at all. They are spending serious money to patent it, license it, and build owned infrastructure around it. The rails for generating and managing AI creative work are being built — and they are being built as property.
It is worth being careful here, because not every patent is a wall. A patent can also be defensive — a way to keep a piece of infrastructure open, free at the low end, standards-based, so that it cannot later be fenced off and turned into someone’s tollbooth. The enclosure a creator should actually worry about is the other kind: the kind that gates the tools and the outputs themselves, and decides who is allowed to be paid for using them.
Set that distinction aside and look at the two pressures together. One says AI creation should not happen. The other says it should happen, but inside someone’s fences. Notice that neither of them describes a world where an ordinary person makes something with these tools, freely, and holds real ownership of what they made. The prestige of old craft and the revenue of large institutions are not coordinating — but they happen to point the same direction, and the independent creator is standing where the two lines cross.
The question that actually has an answer
So stop asking who owns it. That question routes you straight back into copyright — and copyright has already told you the answer depends on proof.
Ask the question underneath it: when this work was finished and went out into the world, was there a verifiable, independent record of who made it and when — a record that travels with the file and does not depend on anyone’s goodwill to stay readable?
That is provenance, and it is a different thing from copyright. Copyright is a claim you file with an office, after the fact, and hope is honored. Provenance is a record — created at the moment a finished work is published, cryptographically signed, carried inside the file, and checkable by anyone against an open standard, without any single platform’s permission.
This is not a new idea. In the art world, provenance has always meant the records created at the time of an event — the dated bill of sale, the gallery invoice, the exhibition catalog — the documents that establish who held a work and when, precisely because they were made then and not reconstructed later from memory. A cryptographic provenance record does the same job for a digital file: it fixes who published it, and when, at the moment it happened, in a form no one can quietly revise afterward.
It is worth being precise about what that record does and does not do. It is not a verdict that a human was creative, and it does not, on its own, sort the human contribution from the machine’s. Think of it the way you think of a notary. A notary does not vouch that your contract is wise, or that your signature is skilled — only that you signed it, that it was you, on that date. A provenance record is that kind of witness for a creative file: it attests that this work, in this exact form, was published by this person at this time. That is evidence — contemporaneous, tamper-evident, identity-bound — and it is the anchor the rest of an authorship case is built around. It is not the whole case. But it is a fixed point, and right now most creators have no fixed point at all.
And look at the price
Set the law aside for a moment and just look at cost.
Registering a copyright is a paid filing — a fee per work, plus the time to prepare the application, plus the wait for it to process. For a creator releasing one carefully built album a year, that is manageable. For a creator making dozens or hundreds of pieces, it is a real expense and a real bottleneck. And for AI-assisted work, you are now also paying to assemble the human-authorship documentation the Office requires — documentation you mostly don’t have.
Provenance runs the other way. A cryptographic record of authorship can be created at the moment a finished file is published, automatically, for free or for a cost measured in fractions of a cent. No application. No wait. It scales to one piece or ten thousand without the price changing in any way you would notice.
So the choice is not only between protection that is conditional and a record that is automatic. It is between a slow, paid, per-work filing and an instant, near-free record that travels with everything you make. The economics point exactly where the law already pointed.
The burden is yours
Here is the hard sentence, and then the reason it is actually good news.
The burden of documenting your authorship is now yours. Not the platform’s. Not the Copyright Office’s. Not the AI company’s. Yours.
That sounds like a punishment. It is the opposite.
A record you carry yourself cannot be taken away from you. A platform can delete your account and your “proof” with it. A company can change its terms. But a cryptographic record of authorship that lives in your file, that you hold, that anyone can verify against an open standard — that is not trapped inside any single platform that can lock you out. It cannot be revoked by one company’s policy change or one company’s shutdown. That is what makes it genuinely yours.
Carrying your own evidence is not the consolation prize. It is the most durable protection available in this environment — the one no single company can switch off.
What becomes possible after
There is one more reason this matters, and it is the hopeful one.
Right now, an independent creator has two options when AI companies want training data: get scraped without consent and maybe sue afterward, or stay invisible. Both are bad. There is no third option — not because no one wants it, but because no one has built the infrastructure for one.
Provenance is that missing infrastructure. If creative work carries a verifiable record of authorship and the creator’s terms, then “do not train on this” becomes a real, machine-readable signal instead of a wish — and so does “you may train on this, here is my price.” You cannot enforce a preference no one can see; provenance is what makes the preference legible in the first place. Enforcement still depends on adoption and, eventually, on law. But the signal has to exist before any of that is possible, and today it doesn’t.
This is also where provenance still has room to grow. Today, a provenance record can establish that a work was made, by whom, and when. The harder and more valuable step — the one the field is already moving toward — is provenance that also records how: the process itself, captured as it unfolds rather than asserted afterward, under the creator’s own control and held as the creator’s own evidence, not a platform’s surveillance. That is the direction this has to go. What exists now is step one of a longer build, not the finished answer.
Follow the path to the end
It is worth asking where the enclosure path actually leads.
If the infrastructure for AI creative work keeps getting built as owned property, and the independent creator stays locked outside the revenue, follow that forward. The creator pays every month for the tools. The creator supplies the taste, the labor, the iteration, the judgment. And the creator earns — what? Pennies, if that. Maybe nothing.
Meanwhile, look at the scale of what is being built. ABI Research projects the number of data centers operating worldwide will rise from roughly 8,800 in 2026 to over 12,000 by 2035, with North America alone adding well over a thousand new facilities. The firm expects AI workloads to overtake legacy workloads in active data center capacity by 2031. I am not claiming those buildings exist to extract from creators — most of that capacity serves ordinary enterprise computing. The point is narrower and harder to argue with: an industrial layer is being built, at a cost of many billions, because serious people expect a serious return. The only open question is whether the creator is anywhere in that return.
So here is the question that should stop everyone, on every side of the AI argument:
If a small creator cannot earn a real income from the work, how does that creator justify spending hundreds of dollars a month on the tools to make it?
There is no good answer to that. A system where the cost of creating scales up, the infrastructure scales up, the investment scales up — and the creator’s income scales toward zero — is not a creative economy. It is an extraction machine with a hobby at the bottom of it.
This is the path provenance is meant to interrupt. Not because a verifiable record of authorship is a complete economic answer — it is not. But because nothing else becomes possible without it. You cannot be paid for work you cannot document is yours. You cannot license what you cannot evidence. You cannot opt in, on your own terms, to a market you have no standing in. Provenance is the floor a creator has to be standing on before any fair economics can be built on top.
Put the fight down, pick up the proof
None of this is anti-AI. None of it is pro-AI. It is true no matter which side of that argument you started on.
And none of it says copyright stopped mattering. For AI-assisted work it still does — the human contribution is real, and it is worth claiming. The point is different. The point is that the debate itself — is it art, who owns it, should it be banned, should it be protected — has become the entire center of gravity, and it is absorbing enormous energy from the people who can least afford to spend it.
That energy has somewhere better to go. There is a path that is cheaper than a filing fee, faster than a registration wait, and does not require winning a single argument with anyone: hold the record of what you made. It already exists. It scales to everything you make. And once it is in your hands, no policy change and no shutdown can take it back.
So stop pouring yourself into the fight over who owns it. The better way is not theoretical and it is not years away — it has already been laid out. Start holding the record of who made it, when, and how. The burden is yours now — and that is the first thing in this entire mess that no one can take away from you.
—Pat Casino
Sources
ABI Research, “How Many Data Centers Are There and Where Are They Being Built?” (July 16, 2024; page updated March 3, 2026). https://www.abiresearch.com/blog/data-centers-by-region-size-company
U.S. Copyright Office, Copyright and Artificial Intelligence, Part 2: Copyrightability (January 2025).
Thaler v. Perlmutter, 130 F.4th 1039 (D.C. Cir. 2025); cert. denied, No. 25-449 (U.S. Mar. 2, 2026).
Allen v. Perlmutter, No. 1:24-cv-02665 (D. Colo.) — pending.