AI Art and Copyright: Who Owns What You Generate?

Last updated: October 2025

You spend 20 minutes crafting the perfect Midjourney prompt. The AI generates a stunning image. You want to use it on your website, sell it as a print, or include it in a client deliverable.

Who owns it? You? Midjourney? Nobody? The answer depends on where you live, which tool you used, and how much human input you contributed. And the law is still catching up — though it’s moving faster now.

The Current Legal Landscape (February 2026)

United States

The US Copyright Office has been clear on one point: purely AI-generated works cannot be copyrighted. If you type a prompt and AI generates an image with no further human modification, that image has no copyright protection. Anyone can use it.

In January 2026, the DOJ urged the Supreme Court to deny cert in Stephen Thaler’s latest bid to copyright work created entirely by AI — reinforcing the position that AI alone cannot be an “author” under US law. The Supreme Court’s likely refusal to hear the case effectively closes this legal avenue for now.

But the debate isn’t over. In February 2026, an AI artist publicly argued that copyright should apply to more than just humans, reigniting the conversation about whether the current framework is adequate for AI-assisted creative work.

What counts as “sufficient human authorship”:

• Selecting and arranging AI-generated elements into a larger composition
• Significantly modifying AI output (painting over, editing, combining)
• Using AI as one tool in a larger creative process

What doesn’t count:

• Typing a text prompt (even a very detailed one)
• Selecting the best output from multiple generations
• Minor edits (cropping, color adjustment)

The Thaler v. Perlmutter case (2023) confirmed that AI-generated art without human authorship isn’t copyrightable. The Kashtanova case (2023) showed that a graphic novel using Midjourney images could have copyright on the text and arrangement, but not on the individual AI-generated images.

European Union

The EU takes a similar position but with more nuance. Under EU copyright law, works must be the “author’s own intellectual creation.” AI-generated content without meaningful human creative choices likely doesn’t qualify.

The EU AI Act adds transparency requirements: AI-generated content must be labeled as such, which has implications for commercial use. Enforcement is ramping up in 2026.

China

China has been more permissive. A Beijing court ruled in 2024 that AI-generated images CAN be copyrighted if the user made sufficient creative choices in the prompting process. This remains the most creator-friendly ruling globally and hasn’t been overturned.

Practical Reality

In most jurisdictions, the law is unsettled but trending toward a consistent position: pure AI output isn’t copyrightable, but human-directed AI-assisted work can be. Cases are being decided one at a time, and the rules are evolving.

What the AI Tool Terms Say

Each tool has its own terms of service regarding ownership:

Important: Tool terms grant you a license to use the output, but they can’t grant copyright protection that the law doesn’t provide. You can use the image commercially, but you might not be able to stop others from using it too.

The Training Data Problem

The other side of AI art copyright: the training data. AI image models were trained on billions of images scraped from the internet, many of which are copyrighted works by human artists.

Ongoing lawsuits:

• Getty Images vs. Stability AI (using Getty’s copyrighted photos for training)
• Class action by artists against Stability AI, Midjourney, and DeviantArt
• New York Times vs. OpenAI (text, but sets precedent for images)
• Music industry cases that will shape 2026’s AI copyright disputes

These cases could reshape the entire AI art landscape. If courts rule that training on copyrighted images is infringement, AI image generators might need to retrain on licensed data only — which would change output quality and availability.

Adobe’s approach: Adobe Firefly is trained exclusively on Adobe Stock images, openly licensed content, and public domain works. This makes it the safest option for commercial use from a copyright perspective.

Practical Advice for Creators

If You’re Using AI Art Commercially

1. Use Adobe Firefly for anything legally sensitive. The training data is clean, and Adobe offers indemnification for commercial use.
2. Modify AI output significantly. The more human creative input you add, the stronger your copyright claim.
3. Don’t rely on AI art as your sole competitive advantage. If it’s not copyrightable, competitors can use similar images.
4. Keep records of your creative process. Document your prompts, iterations, and modifications. If ownership is ever disputed, this evidence helps.
5. Check your client contracts. Some clients require that deliverables be copyrightable. AI-generated art might not meet this requirement.

If You’re an Artist Concerned About AI

1. Opt out of training data where possible. Stability AI and others offer opt-out mechanisms. Use them.
2. Use tools like Glaze or Nightshade to protect your images from being used in AI training.
3. Follow the lawsuits. The legal landscape is shifting, and artist-favorable rulings could create new protections.
4. Adapt, don’t just resist. Artists who learn to use AI as a tool (while maintaining their unique style and vision) are thriving. CNET’s take — “the only way to stop AI art in 2026 is to make it uncool” — captures the cultural tension, but the practical reality is that AI art tools aren’t going away.

The Bottom Line

AI art copyright is clearer than it was a year ago, but still far from settled. The DOJ’s January 2026 position reinforces that pure AI output isn’t copyrightable in the US. The safest approach: treat AI-generated images as non-copyrightable unless you’ve added significant human creative input. Use AI as a starting point, not a final product. And if copyright matters for your use case, use Adobe Firefly or modify AI output substantially.

The major training data lawsuits will likely produce rulings in 2026-2027 that reshape the landscape further. Until then, proceed with informed caution.

For AI image tools, see our Midjourney v6 Review and Midjourney vs DALL-E vs Stable Diffusion comparison.