© 2024 Corporate and Allied Attorneys. All Rights Reserved.
Sep 16, 2026 | In Blog
Generative AI can produce a logo in seconds. Whether anyone owns that logo is a much slower question.
Imagine you open an AI tool, type a few sentences describing what you want, and moments later you have a logo, a song, an essay or a painting. Who owns it?
These are not academic questions. Businesses are already using generative AI for logos, marketing copy, product images and jingles, often without pausing to ask whether they own the results. If the answer is no, the work falls into the public domain: a competitor can copy it freely, and there is nothing to license, assign or enforce. That is a commercial problem long before it becomes a legal one.
Two of the four answers can be dealt with quickly. The AI itself cannot own the copyright, because the law does not recognise artificial intelligence as a legal person. Copyright belongs to authors, and an author has to be someone the law can treat as a holder of rights — someone who can sue, license, sell or bequeath. A software model can do none of those things.
Nor should the developer own the output simply for having written the software. That would be like a pen manufacturer claiming the novels written with its pens. Most developers agree: the terms of service of the major AI tools generally assign whatever rights exist in the output to the user. But a contract can only transfer rights that exist in the first place, and that is precisely what is in doubt. Which brings us to the hard question — does the user who typed the prompt automatically own the result? This is where existing copyright law struggles.
Copyright protects creative expression; it does not attach to whoever owns the equipment. A photographer owns copyright not because she possesses a camera, but because she chooses the composition, the lighting, the moment and the framing. The camera records; the photographer decides. Courts settled that point well over a century ago, when photography itself was dismissed as a merely mechanical process incapable of authorship. Artificial intelligence, however sophisticated, is the same kind of thing: a tool in someone’s hands.
The leading decision is Thaler v Perlmutter. Stephen Thaler applied to register an artwork called A Recent Entrance to Paradise and told the U.S. Copyright Office plainly that the image had been generated autonomously by his computer system, with no creative input from him at all.
The Office refused registration for want of human authorship. In March 2025 the U.S. Court of Appeals for the D.C. Circuit agreed, holding that human authorship is required as a matter of statutory law, and in March 2026 the Supreme Court declined to hear the case. The American position is therefore settled: copyright does not protect works generated solely by artificial intelligence, no matter how impressive the result.
But Thaler was an easy case, and easy cases resolve little. It tells us what happens when a machine works entirely alone. It says nothing about the far more common situation, in which a person and an AI tool work together — which is how almost everyone actually uses these systems.
The U.S. Copyright Office addressed this directly in January 2025, in Part 2 of its report on Copyright and Artificial Intelligence. Its conclusion was that AI-assisted works can be protected, but only so far as a human supplied the expression. Prompts alone are not enough, however many of them there are and however carefully they are refined, because prompts “may reflect a user’s mental conception or idea, but they do not control the way that idea is expressed”. What can be enough is creatively selecting and arranging AI-generated material, substantially modifying the output, or feeding in the person’s own existing work — and even then, protection covers only the human-authored part, not the AI-generated material underneath it.
This still leaves a gap. We are told that prompting alone is too little and that substantial editing may be enough, but not where the line falls in between; the Office has said only that the question must be answered case by case. One registration gives some sense of the scale involved. In January 2025 the Office registered an image titled A Single Piece of American Cheese, where the artist had used her tool’s inpainting feature to make dozens of separate, deliberate edits to the composition. Even then, the registration covered her selection, arrangement and modifications — not the raw AI output. That is a demanding standard, and most everyday uses of AI will fall well short of it.
Rather than adopting an all-or-nothing approach, copyright law should recognize a human creative control standard. Under this approach, copyright should not depend simply on whether AI was used but on the quality and significance of the human contribution. In practice, that standard would turn on questions a court can actually assess: who decided the structure and composition of the work; whether the person selected, rejected and rearranged what the tool produced or simply accepted the first acceptable result; how much of the finished work reflects deliberate human choice; and whether the person can explain their creative decisions rather than merely repeat their instructions.
This approach preserves the principle that copyright rewards human creativity while acknowledging that creativity increasingly occurs through technologically advanced tools. It also fits the pattern of how copyright has always absorbed new technology. Cameras, synthesisers, photocopiers and design software each provoked the same anxiety, and in each case the law eventually returned to the same question: what did the human being decide?
It is worth separating two questions that are constantly confused. Everything above concerns the output — who owns what an AI tool produces. A second and equally contested question concerns the input — whether it is lawful to train an AI model on copyrighted works in the first place. The headline lawsuits are almost all about the second question, not the first.
Both are still unfolding. In Getty Images v Stability AI, the English High Court gave judgment on 4 November 2025. Getty had to abandon its principal training claim mid-trial because it could not show that the training had taken place in the United Kingdom, and the court held that a model’s weights are not a “copy” of the training images, so distributing the model in the UK was not secondary infringement. Getty succeeded only on a narrow and largely historic trade mark point about watermarks appearing in outputs. In the United States, the artists’ class action in Andersen v Stability AI remains in discovery, with trial listed for 2027; claims for direct and induced infringement have survived dismissal, while the DMCA claims did not.
Neither case answers who owns AI output. Taken together, though, they show a legal system being asked questions its drafters never contemplated — and answering them slowly.
Everything so far is American and English law. Neither binds a Ghanaian court, and readers here will reasonably ask what our own law says.
The short answer is that the Copyright Act, 2005 (Act 690) does not mention artificial intelligence at all — but it does not need to in order to reach the same destination as Thaler. Section 76 defines an author as “a person who creates a work”. Section 1(2)(a) requires a work to be original in character, and section 1(4) explains that a work is original “if it is the product of the independent effort of the author”. Section 12(1) then protects the work for the life of the author and seventy years afterwards — a term that only makes sense if the author is a human being capable of dying. Read together, these provisions assume a human author from beginning to end. A work generated entirely by a machine, with no human creative effort behind it, would struggle to satisfy any of them.
What Act 690 does not tell us is the harder question this article has been circling: how much human effort counts as “independent effort” when the effort consists of directing a machine. That gap is the same one the U.S. Copyright Office has been wrestling with. Until a Ghanaian court or Parliament addresses it, businesses here are operating on assumption rather than authority.
The United Kingdom took a different route decades ago. Section 9(3) of the Copyright, Designs and Patents Act 1988 provides that where a work is computer-generated and has no human author, the author is “the person by whom the arrangements necessary for the creation of the work are undertaken”, with a shorter fifty-year term. It was a striking provision for 1988, and it has been little used since. Tellingly, in March 2026 the UK Government proposed removing it altogether for want of evidence of its continuing value, while confirming that works created with AI assistance should still be protected. However, section 9(3) of the Copyright, Designs and Patents Act, 1988 remains in force pending any legislative reform. The one country that came closest to giving purely machine-made works their own copyright is now stepping back from the idea.
Artificial intelligence should not be treated as the author of creative works. Instead, copyright law should protect people when they make real creative contributions using AI as a tool. As AI continues to develop, the law must adapt while keeping its focus on protecting human creativity. The law will have to adapt as the technology does, but the question at its centre should stay the same: what did the human being actually decide?
For now, anyone using generative AI commercially should assume the raw output is not protected, keep a record of the creative choices and edits they make to it, and treat what the tool produces as a starting point rather than a finished product. Legally speaking, the value is in what you do next.
This article is for general information only and does not constitute legal advice. For advice regarding a specific legal matter, please contact Corporate and Allied Attorneys.
© 2024 Corporate and Allied Attorneys. All Rights Reserved.