US Copyright Office rules AI-generated Midjourney images can’t be copyrighted

If you prompt an AI to generate a batch of stunning comic book panels, do you actually own them?

The US Copyright Office just weighed in, and the answer is basically: you own the layout and the text, but the individual images belong to no one.

Reuters reports that the office has partially canceled the copyright registration previously granted to Kristina Kashtanova for her comic book Zarya of the Dawn. The agency determined that Kashtanova is indeed the author of the story and the specific arrangement of elements, but the artwork generated by Midjourney cannot be copyrighted because it lacks human authorship.

The ghost in the prompter

The whole saga started last fall when Kashtanova received an initial copyright certificate. At the time, the Copyright Office had no idea the artwork came from a text-to-image generator. Midjourney was only mentioned once on the cover page across an 18-page filing, which officials concluded did not constitute proper disclosure.

The agency only launched an investigation after seeing Kashtanova's celebratory posts on social media. In its final ruling, the office cited historical precedents where non-human authors, such as spiritual entities and the infamous monkey selfie, were denied copyright protection.

The office dug into the mechanics of diffusion models, noting that Midjourney turns prompts into tokens and pulls from training data in ways users cannot predict. That lack of predictability, the letter argued, fundamentally separates AI generators from traditional artist tools. Even Kashtanova’s minor manual edits on certain panels were dismissed as too imperceptible to qualify for protection.

Kashtanova’s attorney, Lindberg, pushed back hard against the decision, arguing that the Copyright Office fundamentally misunderstands how prompt engineering works. Lindberg claimed that directing Midjourney is less like rolling dice and more like directing a complex system, arguing that AI-assisted art will eventually be treated like photography as legal precedents mature.

Why it matters

This ruling draws the first real regulatory boundary for generative AI in creative industries. For creators and businesses relying on Midjourney, the takeaway is stark: raw generated outputs live in the public domain, meaning anyone can freely rip your generated graphics without infringing your copyright.

If you want legal protection today, you cannot rely on prompt engineering alone. You have to weave those assets into a larger compilation, provide original written narratives, or introduce substantial manual edits that clearly demonstrate human craft.

The Copyright Office took a clever route by focusing on unpredictability, effectively ruling that telling a machine what to imagine is not the same as executing the art. However, this stance creates immediate tension with real-world workflows. As prompting tools gain deterministic control over pixels, the line between an unpredictable algorithm and a high-tech camera lens will get much harder for regulators to defend.