The ghost in the machine: can you truly own an ai-generated game character?

Imagine this: you have spent weeks perfecting a prompt. You have gone through thousands of iterations on Midjourney or DALL-E to create the perfect protagonist for your next RPG—a cybernetic knight with a very specific, haunting aesthetic. You put him on the cover of your game, build your entire marketing campaign around him, and even start selling figurines.

Then, a month after launch, a rival studio releases a mobile game using the exact same character. You call your lawyer, ready to sue for copyright infringement, only to hear the words that every developer dreads: “You cannot sue. You do not own the copyright to that character.”

Welcome to the new frontier of gamedev law. As artificial intelligence becomes an integral part of the creative pipeline, the question of intellectual property (IP) ownership is no longer theoretical—it is a significant business risk.

To understand why you might not own your ai knight, we have to look at the very foundation of copyright law. In almost every jurisdiction—from the US Copyright Office to the European Union Intellectual Property Office (EUIPO)—copyright requires “human authorship.”

Historically, the law views tools like cameras, paintbrushes, or even Photoshop as extensions of the human hand. In those cases, the artist has “creative control” over every stroke or frame. However, the law currently views generative ai as a “black box.” When you type a prompt, you are giving an instruction, but the machine is making the creative decisions regarding lighting, composition, and fine details.

If the machine makes the final choices, the law does not recognize a human author. This means the output is born directly into the public domain, meaning anyone can use it, and no one can own it.

The most important legal precedent we have today is the case of the comic book “Zarya of the Dawn” by Kris Kashtanova. Kashtanova used Midjourney to illustrate a 18-page comic book and attempted to register the copyright.

The US Copyright Office issued a landmark ruling:

• The arrangement is protected: Kashtanova was granted copyright for the selection, coordination, and arrangement of the images and the text.

• The images are not: The individual images generated by Midjourney were denied protection. The office ruled that because the artist could not predict or control the specific output of the ai, the machine—not the human—was the creator.

For a game developer, this is a warning. You might own the “game” as a whole, but you may not own the individual characters if they are raw ai outputs.

A common argument from developers is: “But I spent 100 hours perfecting the prompt! That is my creative effort!”

Legally, this is a very difficult position to defend. Courts have compared prompting to a patron telling a master painter: “Paint me a sunset with a lonely tree.” The patron has the idea, but the painter (the ai) has the creative expression. Copyright protects the expression, not the idea.

Unless you can prove that you had specific, granular control over every pixel in that image—which is technically impossible with current diffusion models—your prompt is seen as a suggestion, not a creative act worthy of ip protection.

If you are operating a studio in North Macedonia, the legal framework is particularly traditional. According to the Macedonian Law on Copyright and Related Rights (ZAPSP), an author is defined strictly as a natural person (физичко лице) who has created the work.

In the context of our domestic legislation:

• Article 12 of ZAPSP: This article clearly states that the author is the person who created the work. Since an algorithm is not a natural person, it cannot be an author.

• The risk for local studios: If a Macedonian studio generates its assets purely through ai, those assets lack domestic protection. This means if a competitor in Macedonia copies your ai-generated characters, your legal standing to sue under the domestic copyright law is incredibly weak.

• EU alignment: Macedonia is aligning its laws with the EU Digital Single Market directives. The current European stance is that “originality” must reflect the author’s own intellectual creation. Pure ai outputs do not yet meet this threshold.

Does this mean you should stop using ai? Absolutely not. It means you need a legal hygiene strategy. To claim copyright, you must inject significant human creativity back into the process.

• The overpainting method: Use ai to generate the base concept. Then, have a human artist manually repaint, edit, and transform the image in Photoshop or Blender. The more manual changes you make, the stronger your claim to authorship becomes.

• Derivative works and documentation: Document your process. Keep the original ai output and the multiple versions of human edits that followed. If you can show an “evolutionary trail” where a human made the critical aesthetic decisions, you have a much better chance of protecting that character.

• Hybrid pipelines: Use ai for things that do not need to be owned—like background textures or minor NPC variations. But for your Hero characters, mascots, and logos, make sure they are hand-drawn.

There is another side to this coin: infringement. Ai models were trained on billions of images, often without the consent of original artists. If your ai-generated character looks “suspiciously similar” to a famous artist’s style or a character from an existing franchise, you could be the one getting sued.

Even if you cannot own the ai art, you can still be held liable for using it if it infringes on someone else’s rights.

In the gold rush of technology, it is easy to forget the legal bedrock. Using ai is a powerful force multiplier for indie studios, but it must be used strategically. A character that cannot be owned is an asset that has zero value on a balance sheet. By ensuring that your key ip has a clear, human-driven chain of title, you are protecting the future of your studio.