Your Notes Are Yours — What the AI Wrote May Belong to Nobody

Your note is a human-authored work. The summary a model wrote from it may not be a work at all. The U.S. Copyright Office will not register material whose traditional elements of authorship were produced by a machine — and both OpenAI and Anthropic assign you their interest in the output only "if any."

That is not a forecast. It is what six primary documents say in their own words: the Copyright Office's Compendium of U.S. Copyright Office Practices, its 2023 statement of policy on AI-generated material, a 2025 federal appellate opinion, the Supreme Court docket that ended its appeal, and the published terms of the two assistants most people point at their notes.

The Compendium states the underlying rule in a single sentence: "The U.S. Copyright Office will register an original work of authorship, provided that the work was created by a human being."¹ Everything downstream of that sentence is a consequence of it. The registrability of a generated summary. What a vendor can actually give you. What survives when a human edits the output.

The interesting part is not that the rule exists. It is how narrow it turns out to be.

The floor under the note you typed

A note you wrote by hand has a legal floor under it: the Copyright Office registers an original work of authorship on the condition that a human being created it. That condition is the whole rule. It is not a formality, not a filing requirement, and not new — it predates the typewriter.

The Compendium is blunt about failing that condition. "Because copyright law is limited to 'original intellectual conceptions of the author,' the Office will refuse to register a claim if it determines that a human being did not create the work."² Refuse, not defer.

The authority is nineteenth-century. §306 grounds the rule in Trade-Mark Cases, 100 U.S. 82, 94 (1879), which limits copyright to "the fruits of intellectual labor" that "are founded in the creative powers of the mind," and in Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53, 58 (1884).³

So the journal entry you typed on a Tuesday clears the condition everything downstream depends on: a human being created it. No registration, no notice, no ceremony required.

What follows is about the other file — the one the model produced.

The rule is older than the machine it now governs

The section that says a machine cannot author a work was written about photocopiers and animals, not about language models. It predates the technology it is now quoted against by decades, and its own illustrations are a monkey, an elephant, and driftwood. Search the chapter for "artificial intelligence" and you find nothing.

That matters, because §313.2 is routinely called the Copyright Office's AI rule. "Works That Lack Human Authorship" says: "To qualify as a work of 'authorship' a work must be created by a human being. … Works that do not satisfy this requirement are not copyrightable."⁴

Its clause about machines reads: "Similarly, the Office will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author."⁵

The examples themselves are pre-digital: "A photograph taken by a monkey." "A mural painted by an elephant." "A claim based on driftwood that has been shaped and smoothed by the ocean."⁶ Search all thirty-nine pages of Chapter 300, revised January 28, 2021, for "artificial intelligence" and the count is zero.⁷

The chapter is not silent on the underlying question. It quotes the Register of Copyrights' 1965 report to the Librarian of Congress, which framed the test everything since has circled: whether the "work" is "basically one of human authorship, with the computer [or other device] merely being an assisting instrument, or whether the traditional elements of authorship in the work … were actually conceived and executed not by man but by a machine."⁸

Sixty years old. Still the question.

The document that actually addresses AI is from 2023

The document that applies the human-authorship rule to AI is a 2023 statement of policy in the Federal Register, 88 FR 16190. It is explicit where the Compendium is silent: when a technology receives only a prompt and produces the work, the Office treats the authorship as the machine's, and will not register it.

The Office states the constitutional premise first: "Most fundamentally, the term 'author,' which is used in both the Constitution and the Copyright Act, excludes non-humans."⁹ Then the operative test: "If a work's traditional elements of authorship were produced by a machine, the work lacks human authorship and the Office will not register it."¹⁰

The prompt case is addressed by name. "For example, when an AI technology receives solely a prompt from a human and produces complex written, visual, or musical works in response, the 'traditional elements of authorship' are determined and executed by the technology—not the human user."¹¹

Read that against what you did this morning. You opened a chat window and typed summarize my journal for the last month. That is a prompt. The Office's analogy is the sharpest line in the guidance: "Instead, these prompts function more like instructions to a commissioned artist—they identify what the prompter wishes to have depicted, but the machine determines how those instructions are implemented in its output."¹²

You commissioned it. Commissioning is not authoring.

A court held the machine cannot be the author — and refused to say more

In 2025 the D.C. Circuit held that a machine cannot be listed as the author of a copyrighted work. It then declined to answer the question most readers actually have — whether making and using the tool makes the human its author — because that argument had been waived before the agency.

The holding, written by Circuit Judge Patricia A. Millett for the panel in Thaler v. Perlmutter: "The Creativity Machine cannot be the recognized author of a copyrighted work because the Copyright Act of 1976 requires all eligible work to be authored in the first instance by a human being."¹³

The reservation sits in the same paragraph. "Given that holding, we need not address the Copyright Office's argument that the Constitution itself requires human authorship of all copyrighted material. Nor do we reach Dr. Thaler's argument that he is the work's author by virtue of making and using the Creativity Machine because that argument was waived before the agency."¹⁴

That second sentence is the one to sit with. I built the tool and I ran it, so the output is mine is exactly the argument a person running a model over their own notes would make — and the court expressly did not reach it. It was not decided and not lost on the merits. It was waived below, and left open.

On March 2, 2026, the Supreme Court declined to hear the appeal: the petition for certiorari in No. 25-449 was denied.¹⁵ A denial of certiorari is not a ruling on the merits — it decides nothing about whether the D.C. Circuit was right, and it carries no precedential weight of its own. What it does is leave the D.C. Circuit's decision standing, final and unreviewed, binding in that circuit and persuasive elsewhere.

Declining to hear a case is not deciding it.

Using AI does not forfeit your copyright

Using a model does not forfeit copyright in what you write. The same opinion that refused to name a machine as author says so directly: the human-authorship requirement does not prohibit copyrighting work made by or with the assistance of artificial intelligence. The requirement is about who the author is, not which tools were open.

Millett's words, at pages 18 and 19 of the slip opinion: "the human authorship requirement does not prohibit copyrighting work that was made by or with the assistance of artificial intelligence. The rule requires only that the author of that work be a human being—the person who created, operated, or used artificial intelligence—and not the machine itself."¹⁶

This is the sentence that keeps the subject from collapsing into panic. The case was about listing a machine in the author field of a registration application. It was not a holding that AI-assisted writing is unprotectable, and reading it that way inverts it.

The line runs between who authored the expression and what was in the room.

Two vendors, the same two words

OpenAI and Anthropic both address output ownership in their terms, and both hang the promise on the same two words: if any. OpenAI says you own the Output and assigns its interest if any. Anthropic never says you own outputs; it assigns only its own interest, if any.

OpenAI's Terms of Use, effective January 1, 2026, under "Ownership of content": "As between you and OpenAI, and to the extent permitted by applicable law, you (a) retain your ownership rights in Input and (b) own the Output. We hereby assign to you all our right, title, and interest, if any, in and to Output."¹⁷

Anthropic's Consumer Terms of Service, effective October 8, 2025, §4, is drafted differently. On inputs: "As between you and Anthropic, and to the extent permitted by applicable law, you retain any right, title, and interest that you have in the Inputs you submit."¹⁸ On outputs: "Subject to your compliance with our Terms, we assign to you all of our right, title, and interest—if any—in Outputs."¹⁹

Set them side by side and the difference is structural, not stylistic. OpenAI states affirmatively that you own the Output. Anthropic makes no such statement — the phrases own the Output, you own the Outputs, and own the Outputs appear nowhere in its Consumer Terms.²⁰ It assigns its own interest and acknowledges whatever interest you already have. Two drafting teams, the same hinge, different promises.

Neither clause is a trick. Both are, on their face, generous: each gives you everything the company has. The problem is upstream of both.

An assignment transfers what the assignor holds, and if nothing in the output is protectable because no human authored it, the assignment moves nothing. The "if any" is accuracy, not evasion.

OpenAI is candid about a second hole in the same section: "Due to the nature of our Services and artificial intelligence generally, output may not be unique and other users may receive similar output from our Services. Our assignment above does not extend to other users' output or any Third Party Output."²¹

What you can own, and how

There is a documented way back in. The Copyright Office says a human may select or arrange AI-generated material creatively enough that the resulting whole is an original work of authorship. Protection then covers only the human-authored aspects — the arrangement, the edits, the writing around it. Not the generated material itself.

The guidance is explicit: "For example, a human may select or arrange AI-generated material in a sufficiently creative way that 'the resulting work as a whole constitutes an original work of authorship.'"²² And the limit on what that earns you: "In these cases, copyright will only protect the human-authored aspects of the work, which are 'independent of' and do 'not affect' the copyright status of the AI-generated material itself."²³

So the path is not a trick of phrasing. It is real work, done by you, on top of the output — choosing what survives, ordering it, cutting it, writing the sentences that make it argue something. The raw material stays where it was. Your layer is yours.

One procedural fact belongs here. If you ever register a work, disclosure is mandatory: "applicants have a duty to disclose the inclusion of AI-generated content in a work submitted for registration and to provide a brief explanation of the human author's contributions to the work."²⁴

You cannot quietly register the machine's half.

What this changes about how you keep notes

The practical consequence is small and specific: keep the original. A note you wrote is the part of the pair with a legal floor under it. The summary is derived, regeneratable, and possibly unownable. Store both, label which is which, and never let the derived file become the only copy.

This compounds a discipline this blog has argued before, on accuracy grounds. The summary is a view, the note is the source makes the case for keeping the original so a wrong summary stays reversible. Registrability adds a second reason: the original is the part of the pair whose authorship is not in doubt. One habit, two justifications.

Three moves follow, and none of them requires a lawyer.

1. Keep the pair. note.md and note.summary.md in the same folder. The generated file is a view, never the record.
2. Mark what the machine wrote. Provenance labeling was worth doing for accuracy; it is also how you identify, later and without guessing, which passages the Office would not register.
3. Know which direction the terms run. Output ownership is one relationship; what the vendor may do with your input is another, under different clauses. The training-data question is separate.

None of this is defensive crouching. It is the filing hygiene that makes notes useful in five years, with one added fact: the file you typed is doing legal work the generated file may not be able to do.

What this piece is not

This is reporting on primary sources — the Copyright Office's own guidance, a federal appellate opinion, and two published sets of vendor terms. It is not legal advice. It describes United States law only; the European Union and every other jurisdiction sit outside its scope, and the human-contribution threshold remains unsettled and fact-specific.

Two further limits. Vendor terms are living documents: the OpenAI clause quoted here is effective January 1, 2026, the Anthropic clause October 8, 2025, and either can change without notice. And nothing above tells you whether a particular thing you made is protected. These documents describe what the Office will and will not register, which is narrower than ownership and narrower still than what a court would decide.

Where a reader wants a conclusion these six documents do not state, the honest answer is that they do not state it.

Frequently Asked Questions

Eight questions readers actually type, answered from the same six documents. Where a source does not answer a question, this section says so rather than filling the gap. The recurring theme: the Copyright Office describes what it will and will not register, which is narrower than what most people mean by ownership.

If I ask Claude to summarize my journal, do I own the summary? Your journal is yours; a human wrote it. The summary is the open question. The Copyright Office's position: when a technology receives solely a prompt and produces the work, the traditional elements of authorship are "determined and executed by the technology—not the human user,"¹¹ and it "will not register it."¹⁰ Anthropic assigns you its interest in Outputs "if any";¹⁹ it never says you own them.²⁰

Is AI output automatically in the public domain? That overclaims. The verified position is narrower: the Office will not register a work whose traditional elements of authorship were produced by a machine.¹⁰ Unregistrable is the practical effect, but "public domain" is a conclusion these documents do not state in those words. Use the Office's language, not the shorthand.

Didn't the Supreme Court rule on this? No. It declined to hear the appeal: the petition in No. 25-449 was denied on March 2, 2026.¹⁵ A denial of certiorari is not a decision on the merits and carries no precedential weight. The 2025 D.C. Circuit decision stands, final and unreviewed, but the Supreme Court did not endorse it.

Does using AI at all destroy my copyright? No, and the court said so: the human-authorship requirement "does not prohibit copyrighting work that was made by or with the assistance of artificial intelligence."¹⁶ What matters is whether a human contributed the expressive elements. A machine in the workflow is not disqualifying; a machine in the author field is.

How do I keep protection when I use AI? The Office describes the route itself. A human may "select or arrange AI-generated material in a sufficiently creative way" that the resulting whole is an original work of authorship.²² Copyright then protects "only the human-authored aspects of the work,"²³ independent of the AI-generated material's own status. Your editing layer is the protected layer.

Do I have to tell the Copyright Office I used AI? If you register, yes. Applicants "have a duty to disclose the inclusion of AI-generated content in a work submitted for registration and to provide a brief explanation of the human author's contributions to the work."²⁴ It is a duty on the applicant, and it applies to registration rather than to private use.

OpenAI says I own the output — isn't that settled? OpenAI does say it: "you (a) retain your ownership rights in Input and (b) own the Output."¹⁷ The assignment that follows conveys "all our right, title, and interest, if any."¹⁷ A company can only assign what it holds. If no human authored the material, the clause transfers nothing.

Does this apply outside the United States? No. Everything here is United States law and United States agency practice. The European Union and other jurisdictions handle authorship, originality, and machine-generated material under different frameworks that were not researched for this piece. Treat every statement above as US-scoped.

The sentence you typed has a floor under it. The sentence the machine typed may have none. Keep the one you wrote.

If that discipline needs a place to live, mnmnote.com keeps notes as open Markdown on your own device — readable, and yours, without the tool that helped you write them.