You Don’t Own the AI Content You Just Created

A minimalist black silhouette of a human head with the letters "AI" written above it, filled with numerous googly eyes of various sizes against a plain white background.
Intellectual Property

You Don’t Own the AI Content You Just Created

September 18, 2026

If AI Did the Creating, You Don't Own It, and Anyone Can Legally Copy It

Content generated entirely by AI has no copyright protection in the United States. It has no owner, so it sits in the public domain from the moment it's created, and anyone, including your competitors, can copy, resell, or reuse it without asking you.

And here's the part that catches most people off guard: writing a longer, more detailed prompt doesn't fix this. Not yet, anyway.

Not sure which side of the line your content falls on? Jump to the 60-second checker. It sorts your situation into a category and points you to the right section.

Is Your Content Protected? Try the Checker

Answer three questions. The checker sorts your description into one of three categories and tells you which section to read next. It's a router, not a legal opinion. It can't tell you whether your specific work is or isn't copyrightable. Only the Copyright Office or a court can decide that.

Describe your content

Three labels you'll see throughout

Protected A human contributed enough original expression, and the AI use was properly disclosed. Protection covers the human-authored parts.

Public domain The AI produced the expression. Nobody holds a copyright, so anyone can copy it.

Unsettled The law hasn't answered yet. Right now that mostly means the prompt-engineering question in Allen v. Perlmutter.

The Legal Shock: Why You Don't Own It

Public domain

This is public domain the moment you make it. If an AI system produced the creative expression and your role was giving instructions, no one holds a copyright in the result. Not you, not the AI company, not the AI itself. With no owner, nobody can use copyright law to stop anyone else from copying it.

Most people assume that whoever makes something owns it. In U.S. copyright law, that's only true when a human being authored the creative expression. In Thaler v. Perlmutter, the D.C. Circuit held that the Copyright Act requires a human author. The Copyright Office's January 2025 report on copyrightability went a step further and concluded that, with today's tools, prompts alone don't give a user enough control over the output to count as its author.

What "human authorship" means in plain English

Copyright protects original expression that a person created: the words on the page, the brushwork, the arrangement of a layout. It doesn't protect ideas, instructions, or the act of asking for something. Type "a watercolor fox in a snowy forest, moody lighting" into an image generator and you've expressed an idea. The system chose the pose, the brushwork, the light, and a thousand smaller details. The Copyright Office's position is that the system, not you, authored those details, and a machine can't be an author.

Think of commissioning an artist. The Copyright Office makes the same comparison: a prompt works like instructions to a commissioned artist. It says what you want depicted, but whoever decides how to express it is the author. The Office adds that a client can oversee and direct a human artist in a way a user can't oversee today's AI systems, and the system itself can't be the author because it isn't a human.

What "public domain" does and doesn't mean here

Public domain is the practical result for the AI-generated parts: with no copyright, copyright law can't be used to stop copying. But a few things still apply.

  • The tool's terms still bind you. Providers like OpenAI say that, as between you and them, you own the output, and they assign you whatever rights they hold in it. They also warn that outputs may not be unique and that other users may receive similar results. That's a contract with the vendor. It can't create a copyright the law doesn't recognize, and it doesn't bind anyone else.
  • Human-authored parts stay protected. Your own text around an AI image, your edits, and your original arrangement can all carry copyright. Section 2 explains how.
  • Other laws can still apply. A rival who copies your AI-generated logo and uses it to confuse customers can face trademark or unfair-competition claims (section 4). And if an AI output happens to reproduce someone else's protected work, whoever publishes it faces a separate infringement risk.

Three places this bites small businesses

  • A marketer publishes 40 blog posts drafted by an AI tool. A competitor republishes them word for word. For the parts the AI wrote, copyright typically won't support a claim.
  • A shop owner sells a printable planner whose art came straight from an image generator. A rival copies the file and lists it for less. Copyright generally won't help stop it.
  • A startup pays a designer for a logo that turns out to be 100% AI output. The client now "owns" rights that nobody ever held (section 5).

Those are illustrations of the general rule, not a verdict on anyone's specific work. Where AI-generated ends and AI-assisted begins depends on the facts, and that's the next section.

AI-Generated vs. AI-Assisted: The Line That Matters

Every question in this area comes back to one thing: who made the creative choices you can actually see in the finished work? Here's how the three situations break down.

Public domain

AI-generated

The AI produced the expressive content. Your input was instructions, picking a finished output, or cosmetic tweaks.

Examples: a blog post generated from a prompt and published as is; a logo straight out of a generator; an image you re-rolled until you liked it.

Unsettled

Prompt-heavy

Hundreds of prompts, variations, and re-rolls, but no creative changes made outside the tool. The Copyright Office says that's still instructions. A federal court is being asked to disagree.

Closest real-world test: an award-winning image built from 624 prompt revisions, Midjourney's variation tools, and Photoshop cleanup (Allen v. Perlmutter, section 3).

Protected

AI-assisted

A human made original creative choices that show up in the final work: rewriting, redrawing, compositing, arranging. Protection covers those human contributions.

Examples: a book you wrote, with AI images you selected and arranged; your own drawing that AI extended, where your work is still visible; a draft you rewrote line by line.

Director's chair vs. slot machine

Here's a mental model that holds up well. Prompting can feel like a slot machine. You pull the lever, see what comes out, and pull again. Pull it 600 times and keep the best spin, and you've influenced which result you got, but you didn't decide what's in it. Now picture the director's chair: you rewrite the sentences, redraw the shapes, composite the pieces, and choose where everything goes. Those are expressive choices, and they're yours.

The Copyright Office's current view of prompting is close to the slot-machine view. Its report describes revising and resubmitting prompts as re-rolling the dice, and says the final output reflects the user's acceptance of the system's interpretation rather than authorship of the expression. Jason Allen's lawsuit argues the opposite: that iterating with a tool is creative authorship, the same way a photographer's choices make a photograph the photographer's. (The slot-machine and director's-chair labels are ours, not a legal test.)

Three ways a human contribution can count

The Copyright Office's January 2025 report sorts human contributions to AI output into three buckets.

  • Prompts. With today's tools, the Office concluded that prompts alone don't give users enough control over the expressive result. It treats them as instructions that convey ideas. (A sufficiently creative prompt can itself be protected as text, but that doesn't give you rights in what the AI generates from it.)
  • Your own expression as input. If you feed your own drawing, photo, or text into an AI system and it's still perceptible in the output, you keep copyright in what you contributed.
  • Modification and arrangement. If you creatively edit AI output, or select and arrange AI-generated pieces into an original whole, the human-authored changes and the arrangement can be protected. The underlying AI-generated pieces still aren't.

Using AI to brainstorm, outline, proofread, or clean up your own work doesn't reduce protection for what you authored. The Office treats that as assistive use.

There's no magic percentage

Nothing in the law says "30% human" or "50% human." The Office looks at whether a person made original creative choices in the expressive elements, and whether those choices are visible in the final work. Time and effort don't substitute for that. Allen's hundreds of prompt revisions are the example: the Office's position is that all that iteration still produced instructions, not authorship. Each case turns on its own facts, which is why this article describes the framework instead of telling you where your own work lands.

The Office's own examples show what falls short. In Zarya of the Dawn it found the author's tweaks to a character's lips too minor to count. Its Compendium lists things like sharpening or balancing colors as typical technical alterations that don't warrant registration. And edits made with AI-powered features in editing software are analyzed the same way as AI generation, not as ordinary human editing.

Swipe the table sideways to see every column.

ScenarioCopyrightable?Legal status
Fully autonomous AI output, with no human in the loopNoPublic domain
Settled for this scenario: the D.C. Circuit's Thaler ruling stands.
You write a prompt and publish the output unchanged, or with cosmetic tweaksGenerally noPublic domain
Copyright Office position and practice.
Hundreds of prompts, re-rolls, and variations, with no creative edits outside the toolThe Office says no; no court has ruledUnsettled
The closest live test is Allen v. Perlmutter.
An AI draft that a human substantially rewrites, redraws, or recomposesYes, for the human-authored contributionsProtected
When registering, disclose AI-generated content and disclaim anything beyond minimal.
Your own text, drawing, or photo fed into AI and still visible in the resultYes, for your original inputProtected
Your contribution stays yours. The AI's additions don't.
You select and arrange several AI-generated pieces into an original layout or compilationPossibly, for the arrangement onlyProtected
Arrangement only. See Zarya of the Dawn in section 3.
AI used to brainstorm, edit, or proofread work you wrote yourselfYes, the work is yoursProtected
Assistive use doesn't reduce protection for what you authored.
Sources: U.S. Copyright Office, Copyright and Artificial Intelligence, Part 2: Copyrightability (Jan. 29, 2025); Zarya of the Dawn decision letter (Feb. 21, 2023); Thaler v. Perlmutter (D.C. Cir. 2025; cert. denied Mar. 2, 2026). Reflects the law as of September 19, 2026. These rows are categories, not determinations about any particular work.

The Cases: One Settled, One Still Open

Two cases get mixed up constantly because both captions name the same government defendant, the Register of Copyrights, Shira Perlmutter. The facts have nothing in common. One is over. The other could redraw the line.

Thaler v. Perlmutter Settled

The work
"A Recent Entrance to Paradise," an image that Stephen Thaler said his AI system, the Creativity Machine, created on its own.
Human involvement
None claimed. Thaler listed the AI as the sole author and himself as the owner.
The ruling
The D.C. Circuit (March 18, 2025) held that the Copyright Act requires a human author and upheld the Copyright Office's refusal. Citation: 130 F.4th 1039.
The Supreme Court
Declined to hear the case on March 2, 2026 (No. 25-449), without comment.
What it means
Output created autonomously by AI, with no human author, isn't copyrightable. A denial of review isn't a Supreme Court ruling on the merits, but it leaves the D.C. Circuit's decision standing, and it matches the Copyright Office's long-standing practice.

What Thaler didn't decide: how much human involvement is enough. The court didn't hold that AI-assisted work can never be protected. It left the line-drawing to other cases, which is where Allen comes in.

Allen v. Perlmutter Pending

The work
"Théâtre D'Opéra Spatial," a Midjourney image that won the digital art category at the 2022 Colorado State Fair.
Human involvement
By his own account to the Copyright Office, at least 624 rounds of prompt revisions to reach the initial image, then Photoshop cleanup of flaws and AI upscaling.
Copyright Office
His application didn't mention AI. The Office knew about the work from press coverage and asked. It said it could consider registering only his human-authored contributions if he disclaimed the Midjourney-generated material. He refused, and the Review Board affirmed the refusal in September 2023.
Court
Allen v. Perlmutter, No. 1:24-cv-02665 (D. Colo.), filed September 26, 2024, asking the court to set aside the refusal under the Administrative Procedure Act. Cross-motions for summary judgment are fully briefed.
Status
We found no report of a ruling as of September 19, 2026. Check the docket for the latest before you rely on this.
Unsettled

This case could redraw the line. Allen argues that steering a tool through hundreds of choices is authorship. The Copyright Office answers that his prompts described ideas and Midjourney supplied the expression. Whichever way the court rules, a Colorado district court isn't the last word: the decision could be appealed, and a district court ruling doesn't bind other courts. But it would be the first time a judge weighs in on heavy prompting, and this article will change if it lands.

Keep them separate. Thaler answers "can a machine be an author?" (no). Allen asks "how much can a human do with a machine before they become the author?" Only the second question is open.

Two registrations that show where the Office draws the line

Zarya of the Dawn (2022 to 2023). Kristina Kashtanova registered a graphic novel on September 15, 2022. The application didn't mention AI. After a reporter contacted the Office about her public statements, it opened cancellation proceedings. In its February 21, 2023 decision, it replaced the registration with a narrower one covering her text and her selection, coordination, and arrangement of text and AI-generated artwork, and it kept the original effective date. The Office said the single appearance of the word "Midjourney" on the cover was not a clear disclosure. The lesson is twofold: the arrangement can be protected, and the application fields are where AI use has to be disclosed.

A Single Piece of American Cheese (2025). Invoke registered this image on January 30, 2025, after first being refused. The artist used inpainting to build the piece from about 35 AI-generated elements and recorded each decision. The registration covers the selection, coordination, and arrangement of those elements, not the individual pieces, and it has been widely reported as the first of its kind for a single image. The lesson: documented, hands-on creative control is what the Office responded to.

The Trademark Escape Hatch: Why an AI Logo Can Still Be Protected

Protected, under a different law

Not copyrightable does not mean not trademarkable. Trademark law doesn't ask who or what made the mark. It asks whether the mark tells customers where a product comes from, and whether you're actually using it in business.

Here's the good news most people haven't heard. You'll see some sources claim an AI-generated logo can't be trademarked either. The weight of current legal commentary says otherwise. Trademark protects brand identity, not creative authorship, and the USPTO's registration test looks at three things: whether the mark is distinctive, whether it's used in commerce (or you have a genuine plan to use it), and whether it conflicts with marks that already exist. Nothing in that test turns on how the mark was designed.

The USPTO's April 2024 guidance on AI covers how attorneys and applicants use AI tools when preparing filings. It doesn't say AI-made marks are ineligible. And trademark rights belong to whoever uses the mark in commerce, a person or a company. The AI tool never owns anything.

Swipe the table sideways to see every column.

QuestionCopyright answerTrademark answer
Does it require a human author?Yes. Purely AI-generated material isn't protected.No. The test doesn't turn on who or what designed the mark.
What does it protect?Original creative expression in the artwork itself.The logo's job as a brand identifier: telling customers who a product or service comes from.
What's the key test?Human authorship and originality.Distinctiveness, use in commerce, and no likelihood of confusion with existing marks.
How do rights arise?Automatically when a human creates original expression. Registration is optional but generally needed before suing over a U.S. work.Through real use of the mark in business. Federal registration adds nationwide benefits.
Does it stop someone from copying the artwork?Yes, for the human-authored parts, when protection exists.Only where the copying confuses customers about source. It doesn't lock down the artwork as art.
Where do you file?U.S. Copyright Office.USPTO, through Trademark Center. The base fee is $350 per class as of 2026.
Sources: USPTO trademark and Trademark Center pages; U.S. Copyright Office Part 2 report; current law-firm commentary on AI-generated marks (see Primary sources). Reflects the law as of September 19, 2026.

The honest limits

  • Generic-looking designs can fail the distinctiveness test. A stock leaf icon on a candle label is weak. A leaf that customers associate with your brand can work.
  • AI outputs aren't unique. Generators can produce marks that resemble ones that already exist. A conflict can block your registration or expose you to an infringement claim, so run a clearance search before you commit to a design.
  • Trademark won't stop every kind of copying. It stops confusing uses in commerce. It doesn't stop someone from lifting your artwork as clip art or decoration in a non-competing way. That's the gap copyright normally fills, and it's the gap you have with a purely AI-generated logo.
  • Use has to be real. An application needs genuine use, or a good-faith intent to use, and a real specimen. USPTO examination guidance treats digitally created or mocked-up specimens as not showing actual use, so AI-generated mockups presented as proof of use get refused and can create bigger problems.
  • Check the tool's terms. Make sure the license lets you use its output as a commercial brand mark.

One cheap way to strengthen your position: add your own design work. Custom lettering, redrawn shapes, and a distinctive layout can help with distinctiveness, and they create human-authored contributions that may carry their own copyright.

The law here is still developing. Our read reflects current USPTO practice and the weight of law-firm commentary as of September 2026. We found no USPTO rule requiring human authorship for trademarks, but neither the agency nor the courts have addressed AI-made marks head-on. Treat today's practice as current, not permanent.

The Freelancer Trap: Selling Something Nobody Owns

Say you hire a freelancer for a logo, a brand guide, or 20 blog posts. Your contract probably says the work is "work made for hire," or that the freelancer assigns you all rights, and that the work is original. Now suppose what gets delivered is 100% AI output.

An assignment can only hand over what exists. If no copyright exists in the deliverable, there's nothing to assign, and you end up holding nothing you can enforce against a copycat.

The legal exposure usually isn't copyright infringement. It's contract law. A freelancer who promised original, ownable work and delivered raw AI output may have breached the agreement or a warranty in it, which can mean refunds, termination, or damages. Whether a particular contract was breached depends on its exact wording, so treat this as a question for a lawyer, not a rule of thumb.

If you hire freelancers

  • Ask in writing whether they use AI tools, which ones, and for what.
  • Require disclosure of any AI-generated portions, plus records that show the human contributions: drafts, layered files, edit histories.
  • Decide what you're actually buying. If you need something you can enforce, require human authorship in the parts that matter, and clear logos for trademark separately.
  • Check the AI tool's terms for commercial use.
  • Don't let anyone paste your confidential material into a public AI tool without your permission. That's a separate risk from ownership, and it can breach your own confidentiality obligations.

If you're the freelancer

  • Don't sign "sole author" or "entirely original" warranties you can't back up.
  • Disclose your AI use, and price for the human contribution: creative direction, editing, arrangement, revision.
  • Keep your working files. Layers, drafts, and version history are your proof.
  • Follow each tool's terms of use.

Starter language for both sides is in the clause templates in section 8.

What the client thinks they're buying

  • An exclusive asset nobody else can use
  • The power to stop copycats
  • Something they can register
  • A clean chain of ownership from creator to client
Public domain

What 100% AI content actually delivers

  • A file they can use, and so can everyone else
  • No copyright to enforce against copycats
  • A registration application in which the AI-generated parts have to be disclaimed
  • A possible contract dispute with the freelancer who promised original work

If You Sell AI Content Online (Etsy, Amazon KDP, Shopify)

Printables, planners, wall art, templates, clip-art bundles, low-content books, and product photos are easy to make with AI, and easy to copy. Three separate layers of risk apply, and people mix them up.

Layer 1: copyright, the public-domain exposure

Purely AI-generated material carries the same exposure as anything else in this article. A competitor can copy your file and list it under their own shop. And you can't fall back on a copyright takedown for the AI-generated parts, because you don't hold the copyright. Sending a notice that claims rights you don't have can backfire.

Layer 2: platform rules, which are not copyright law

Marketplaces set their own AI rules. Amazon's KDP policy, for example, asks you to disclose AI-generated text, images, and translations, and as we last checked it treats content as "AI-generated" even if you edited it heavily afterward. The Copyright Office looks at substantial human modification differently. So a book can count as "AI-generated" under a platform's definition and still carry copyright in its human-authored parts. Passing one test tells you nothing about the other.

Etsy's Creativity Standards ask you to classify each listing by your role: made, designed, sourced, or handpicked by you. Etsy's help center places seller-prompted AI creations under "designed by a seller," and seller resources say AI use should be disclosed in the listing. Etsy tightened these standards in 2025, so read the current text before you list. Other marketplaces and book platforms have their own rules too. Don't describe AI work as "hand-drawn" or "original artwork" if it isn't.

Layer 3: the AI tool's own terms

Some tools limit commercial use on free plans or by company size, and outputs can resemble existing works. A listing that draws an intellectual property complaint can cost you the listing, or the shop.

What sellers can do

  • Add real human authorship: your own layout, illustration, writing, and arrangement. That's the part you can protect.
  • Trademark the shop name and logo. Even when the products themselves are easy to copy, the brand isn't.
  • Sell more than the file. Service, community, and being first with a recognizable brand are harder to copy than a PDF.
  • Keep records of your creative contributions as you work.
  • Check each platform's current AI disclosure rules before publishing.

Can Trade Secrets Help? (Only for What You Keep Secret)

Trade secret law protects confidential information that has business value because it's secret, as long as you take reasonable steps to keep it secret. For AI users, that can mean a proprietary prompt library, a repeatable workflow, a custom configuration, or unreleased drafts.

What trade secrets do not do: protect the published output. The moment you publish an AI-generated blog post, image, or design, it's no longer secret, and trade secret law has nothing to say about a competitor copying it. This option covers your unpublished prompts and methods only.

  • No protection against independent creation. If a competitor legitimately arrives at a similar prompt or output on their own, secrecy doesn't stop them. AI outputs aren't unique, so that can happen.
  • Secrecy takes work. Use NDAs with contractors, limit who can see your methods, and check whether the tool's data settings let the provider keep or use your inputs.
  • Secrecy can clash with registration. Registering means explaining your human contributions. In Allen, the Review Board noted that he declined to disclose his specific prompts because he called them confidential, so it couldn't consider whether the prompts themselves were creative enough to be protected. Decide early which matters more for each project.

How to Actually Register AI-Assisted Work

Registration doesn't create copyright in your human-authored work. That exists automatically. But in the U.S. you generally have to register before you can sue over infringement of a U.S. work, and timely registration unlocks stronger remedies. If you register work that involves AI, honesty about the AI is the whole game. The Office says it has registered hundreds of works that incorporate AI-generated material, with each registration covering the human author's contribution.

Don't downplay or skip the AI disclosure. Zarya's registration was cancelled and reissued in narrower form after the Office learned from public statements that Midjourney made the images, and the tool's name on the cover didn't count as disclosure. Allen's application didn't mention AI at all, and the Office refused to register the work as submitted because he wouldn't disclaim the Midjourney material. The Office's guidance also warns that a court can disregard a registration in an infringement case if the applicant knowingly gave inaccurate information, and the Office's own letters note that knowingly false statements of material fact in an application are punishable under 17 U.S.C. § 506(e).

  1. Document as you create. Save drafts, layers, prompt logs, and version history, plus notes on what you selected and changed. Build the record while you work, not afterward. Invoke's successful registration leaned on a record of every decision.
  2. Sort human from AI. List what you wrote, drew, edited, composited, or arranged, and what the tool generated.
  3. Use the Standard Application. The Copyright Office's AI guidance says applicants with AI-generated material must use it, because the other forms don't have fields for disclaiming material.
  4. Describe your human authorship in the "Author Created" field, in plain terms, such as text you wrote or your selection, coordination, and arrangement of the elements. The Office's own guidance offers wording along the lines of "selection, coordination, and arrangement of [human-authored content] created by the author and [AI content] generated by artificial intelligence."
  5. Disclaim the AI-generated material. Anything AI-generated beyond a minimal amount goes under "Limitation of the Claim," in the "Other" field under "Material Excluded," with a short description such as "[description] generated by artificial intelligence." You can add detail in the "Note to Copyright Office" field. A brief statement is enough: the Office doesn't require you to list the tools or the exact proportions. If you're unsure how to fill it out, the guidance says you can give a general statement that the work contains AI-generated material, and the examiner will follow up.
  6. Never list the AI tool or its company as an author or co-author.
  7. Already applied or registered without disclosing? If your application is still pending, contact the Copyright Office's Public Information Office and report the omission. If it's already registered, file a supplementary registration. The Office has warned that leaving the record uncorrected can put the registration's benefits at risk.

Check the Copyright Office's current fees and forms before you file. This section describes how the Office's guidance works. It isn't a filing service, and the phrasing above is illustrative, not a script for your application.

AI disclosure clause templates for contracts

These are starting-point examples so you know what to ask for. They are not a substitute for an attorney reviewing your specific deal, your jurisdiction, and your existing contract.

Clause A: for the party commissioning work

AI Use and Disclosure. Before delivering any Deliverable, Contractor will tell Client in writing whether generative AI tools were used to create any part of it, identify the tools used, and identify which parts were AI-generated. Contractor will not represent AI-generated material as solely human-authored. On request, Contractor will provide records, such as drafts, layered files, or edit histories, reasonably sufficient to show the human creative contributions to the Deliverable. Contractor confirms that its use of AI tools complies with their terms of use and will not enter Client's confidential information into any AI tool without Client's written consent.

Clause B: for the freelancer or contractor

AI-Generated Material. Client acknowledges that Deliverables may contain AI-generated material, that copyright protection may not be available for AI-generated material under applicable law, and that Contractor's assignment covers only such rights, if any, as exist in the Deliverables. Contractor makes no warranty that AI-generated material is eligible for copyright protection.

What This Article Isn't

This guide goes deep on the AI-specific wrinkle in a bigger question, which kind of protection fits what you've made, and our comparison covers the full three-way picture: Trademark vs Copyright vs Patent: Which Do You Need?

If you need to actually file, negotiate, or litigate something, the next step is our Intellectual Property Lawyer: Full Guide.

And it isn't legal advice or a filing service: nothing here, including the checker and the clause templates, can tell you whether your work is protected.

FAQ

Can you copyright AI-generated content?

Not if the AI produced the creative expression. Copyright requires a human author, and the Copyright Office's position is that prompts alone don't make you one. Work that combines AI output with your own creative contributions, such as substantial rewriting, original arrangement, or your own inputs, can be protected in the human-authored parts. Whether a specific work qualifies is decided case by case.

Do you own AI-generated images?

In a contract sense, often yes: many tools' terms say you own the output as between you and the vendor. In a copyright sense, a purely AI-generated image has no copyright for anyone to own. So you may be free to use it, but you generally can't stop others from copying it (section 1).

Does AI content go straight into the public domain?

For the AI-generated parts, effectively yes. No one holds a copyright in them, so copyright law doesn't stop copying. Human-authored parts of a mixed work keep their protection, and other laws, including trademark, contract, and unfair competition, can still apply.

Can someone legally steal my AI content?

If it's purely AI-generated, copyright law generally won't stop them from copying it. That doesn't make everything fair game. Copying that confuses customers about who's behind a product can raise trademark or unfair-competition problems, and copying that breaks a contract you have with the copier can be challenged. Your human-authored additions, if any, remain protected.

Who owns the copyright to ChatGPT text?

Often nobody, for text the AI wrote. OpenAI's terms say that, as between you and OpenAI, you own the output and OpenAI assigns you any rights it holds. That's a contract term, not a copyright. If you substantially rewrite or arrange the text, your human-authored contributions can be protected.

Does paying for ChatGPT Plus give me copyright?

No. A paid plan can change what the vendor's terms let you do, such as commercial use, but it can't create a copyright the law doesn't recognize. Copyright depends on human authorship, not on your subscription tier.

Can I use Midjourney images for commercial use without copyright issues?

Often yes, if you follow the tool's terms. Plan level and company size can matter, and free tiers may be restricted, so read the current terms. Using an image is different from owning a copyright in it. Competitors can generally copy a purely AI-generated image, and outputs can resemble existing works, so review what you publish.

Does prompting count as human authorship?

Not on its own, according to the Copyright Office's January 2025 report. It treats prompts as instructions that convey ideas, with the AI system determining the expressive result. Whether very detailed, iterative prompting can ever count is the open question in Allen v. Perlmutter.

Can you copyright a prompt?

Sometimes. The Copyright Office says a sufficiently creative prompt can be protected as text, like any other writing. But that protection doesn't extend to the output the AI generates from it. Prompts you keep private may also qualify as trade secrets (section 7).

How much human editing is needed for copyright?

There's no set percentage. The Copyright Office looks for original creative choices that are visible in the final work, such as rewriting, redrawing, compositing, or arranging. Substantive creative changes are more likely to count than minor cosmetic tweaks. The Office has treated tiny touch-ups and routine technical adjustments like sharpening or color balancing as falling short, and it analyzes edits made with AI-powered features the same way as AI generation. It's decided case by case, so keep records of what you did.

Can you trademark an AI-generated logo?

Generally yes, if it's distinctive, used in commerce, and doesn't conflict with existing marks. Trademark law doesn't require human authorship. The main risks are generic-looking designs, resemblance to existing marks, and a lack of real use (section 4).

Can clients sue freelancers for delivering AI content?

A client's claim would usually be about contract, not copyright: for example, breach of a warranty of originality, or failure to disclose AI use that the contract required. Whether it succeeds depends on the contract's wording and the facts (section 5).

Can competitors legally copy my AI marketing materials?

For purely AI-generated copy or images, copyright generally won't stop them. Your human-written text, original layout, and brand elements may still be protected, and trademark or unfair-competition law can cover confusing copying. Consider trademark protection for your brand name and logo.

Can I copyright Etsy printables made with AI?

It depends on how much human authorship is in the final product. Purely AI-generated art isn't protected. Your own illustration, lettering, layout, and arrangement can be. Platform AI-disclosure rules are a separate layer (section 6).

What are the Amazon KDP AI content risks?

There are two separate ones. On copyright, purely AI-generated text and images aren't protected, so copies are hard to stop. On platform rules, KDP asks you to disclose AI-generated content, and its definition can include content you edited heavily afterward. Read the current policy before you publish.

What if I combine multiple AI outputs into one work?

Your original selection, coordination, and arrangement can be protected, but that protection is limited to the arrangement. It doesn't cover the individual AI-generated pieces, and arranging just two or three elements generally isn't enough. Zarya of the Dawn and A Single Piece of American Cheese are the best-known examples (section 3).

Is AI-generated code protected in commercial software?

The same human-authorship rules apply. Code the AI wrote without meaningful human creative input isn't protected by copyright, while your own code and original arrangement can be. Also watch the licenses on any open-source code a tool reproduces, and consider trade secret protection for unpublished source code (section 7).

What happens if I don't disclose AI use when registering?

Incomplete disclosure is a documented reason for refusal and can jeopardize a registration later. The Office has said it may take steps to cancel a registration when essential information was left out or is questionable, and a court can disregard a registration in an infringement case if the applicant knowingly gave inaccurate information. If you registered without disclosing, look into a supplementary registration (section 8).

Does the Thaler decision mean AI-assisted work can't be protected?

No. Thaler involved a work with no human author at all, and the court didn't rule out protection for AI-assisted work. The Supreme Court's decision not to hear the case isn't a ruling on the merits.

Does this apply outside the U.S.?

Not necessarily. This article covers U.S. law. As of the Copyright Office's January 2025 report, the UK, India, Hong Kong, and New Zealand have statutes that treat the person who arranges for a "computer-generated" work as its author, though it's unclear how they apply to generative AI. A 2023 Beijing Internet Court ruling protected an AI-assisted image, but it isn't binding precedent in China. And on August 31, 2026, India's Copyright Office found the image at issue in Thaler original but refused to register it because the application named an AI system as its author. If you sell internationally, get local advice.

Primary Sources

This article is for educational purposes only and is not legal advice. AI and copyright law is fast-moving and partly unsettled, and cases like Allen v. Perlmutter could change parts of it. Consult a licensed intellectual property attorney about your specific situation.

Leave Comment

Your email address will not be published. Required fields are marked *

Reach the Editor
AdvoraHQ

AdvoraHQ Editorial

Online

Welcome to AdvoraHQ. We decode complex financial concepts—from tax strategies to market investing—using strictly primary sources and deep research.

Got a specific question, a topic request, or feedback on our research? We'd love to hear from you.

Email the Editor