The power system behind a headline, before it shows up in your work, wallet, rights, or community.
BRIEFING BOX
Core mechanism: The human authorship requirement. Copyright needs a person to make the creative choices, and prompting an AI does not count.
Main risk: A business that builds its brand on AI can’t legally stop anyone from copying it, and still carries the liability if the output infringes.
Decision room: Federal courts and the U.S. Copyright Office, not Congress.
Time horizon: 2026–2028.
Who should care: Solo founders and individual creators, small business owners, in-house marketing and brand teams, agencies and freelancers, investors and acquirers who value brand IP.
What to watch: The Third Circuit’s ruling in Thomson Reuters v. Ross, the basket of training cases in California, Copyright Office registration decisions on AI works, any federal AI bill that reaches a vote, the November midterms.
My call: No binding federal law on AI ownership or creator compensation by the end of 2026; this stays a courts-and-licensing fight (70%). Full reasoning below.
You found a way to keep up. You used AI to build the logo, the website, the product photos, and the words that sell what you make, and you did it for a fraction of what an agency would have charged. The story you heard is that AI is a productivity gift for small business, and that part is true. What no one mentioned is that the work it made for you is, in the eyes of the law, authored by no one, which means it is yours to use and anyone’s to take. [More on The Death of Authorship]
The thing you built with AI to stay competitive is the one thing your competitor is free to copy, because the law says a machine’s work has no owner.
Almost 9 in 10 small businesses now use AI somewhere in their operation, up from about a third in 2023. The most common job they hand it is making things: marketing copy, social posts, product descriptions, logos, websites, the visual identity a business runs on. Owners who make the switch routinely report cutting their branding and content costs by half or more. You traded an agency or a freelancer for a monthly subscription, got a logo and a landing page in an afternoon, and put them to work. That was the smart move. It was also the move that left you owning nothing.
The company selling you the tool makes its money on the subscription, not on your logo. Whether you can defend that logo in court is your problem, and nothing about your inability to own it costs the seller a dollar. So the tool keeps getting better at producing finished-looking work, the price keeps falling, and the question of who owns the result goes unanswered, because answering it was never the seller’s job. The largest companies have already handled their version of this. They have lawyers who document the human edits, designers who rework the machine’s draft until it clears the legal bar, and brand assets old enough to predate the whole problem. The small business that bought AI to replace the lawyer and the designer is the one with no way to close the gap.
Here is the hinge the whole story turns on, and it has a plain name: sufficient human control over the expressive elements.
The U.S. Copyright Office settled on that phrase in January 2025. Copyright protects work a person creates. When a machine makes the expressive choices, the layout, the colors, the shapes, the words, there is no human author, and with no human author there is no copyright. The Office was direct about prompting. Typing a description and accepting what comes back, even after dozens of tries, does not clear the bar, because there is a gap between what you asked for and what the machine decided to produce. You steered it. You didn’t author what it made. And copyright follows the author.
So the real fight, the one happening inside the Copyright Office and a stack of federal courtrooms right now, is over a single question: when you make something with a machine, who owns it? For the machine’s part of the work, the answer right now is no one.
That answer is not new. It comes from a rule called the human authorship requirement, and 2025 was the year it stopped being trivia. A computer scientist named Stephen Thaler tried to register a picture his AI made, listing the machine as the author. The Copyright Office said no, a federal appeals court agreed, and in March 2026 the Supreme Court declined to touch it. An artist named Kristina Kashtanova got a harder lesson: she made a graphic novel with AI images, and the Office protected her writing and the way she arranged the book while refusing protection for the images themselves. She had done the work. The law still told her the machine’s part belonged to no one. What was a curiosity when the author was a monkey with a camera became a business problem when the author was the tool half the country now runs on.
When the people in charge all do the same surprising thing at once, that is worth reading. In May 2025 the Copyright Office published a report concluding that much of the AI industry’s training on copyrighted work went beyond fair use. The President fired the head of the office the next day. She sued, an appeals court put her back, and she testified before the Senate this past May. You do not remove the referee one day after a call unless the call mattered to someone with power.
One honest complication, because you will hit it the moment you talk to a lawyer. You are not defenseless. A logo can often still be registered as a trademark, because trademark law asks whether a mark is distinctive and used in commerce, not whether a human drew it. And if you edit, arrange, and rework the machine’s output enough, the human parts can earn copyright. That is real, and it is the right advice. The catch is what a trademark does and does not do. It stops a competitor from using a confusingly similar mark to pass their business off as yours. It does not stop them from copying the image itself for their own, different business. For the copyright that would stop the copying, you need the human authorship you skipped when you let the machine do the creating. Which is the whole problem in one line: the cheaper and faster you let AI build your brand, the less of it you can defend.
The company sells you a tool that produces work the law says has no author. It collects your subscription whether you can own the result or not. And it leaves the ownership gap sitting on the smallest businesses, the ones who bought the tool because they could not afford the lawyers and designers who know how to close it.
The savings are real, and the exposure is the price of them. The one lever that matters is human authorship. If AI builds your brand from start to finish, you own none of the machine’s work and can be copied at will. If a person makes the meaningful creative choices and you keep the receipts, the prompts, the drafts, the record of who changed what, you can own the result. The cost of protection is the human hour you used AI to remove. So decide which assets are worth that hour and put the hour back into those.
If you make content for clients or an employer, the deliverable now carries a question it did not carry before: who owns this, and can we defend it? An agency handing over AI work it cannot copyright is handing over an asset the client cannot protect. The new line in the contract is who carries that risk. Settle it in writing before the work ships, not after a competitor reuses it.
The civic problem here is that the rule stays invisible until it fails. No one tells you at checkout that the logo you just generated cannot be copyrighted. The remedy that matters is disclosure and documentation: knowing, before you build on an asset, whether it is human-authored enough to own. The people who lose are the ones who find out during a dispute.
Main Street adopted AI fastest where the budgets are thinnest. The shop that generated its sign, its menu, and its ads did the rational thing, and ended up with a brand it cannot defend against the franchise down the road that can. Two questions separate a safe asset from an exposed one: did a person make the creative choices, and can you prove it? If the answer is no, that asset is on loan from the public domain, and anyone can take it back.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.