Are AI Generated Images Copyrighted

The Current Legal Position

The US Copyright Office has maintained since 2023 that works generated by AI without meaningful human creative control cannot receive copyright registration. This position was tested in the Thaler v. Perlmutter case, where computer scientist Stephen Thaler sought copyright protection for a visual artwork called "A Recent Entrance to Paradise" created entirely by his AI system, DABUS. The district court ruled against Thaler in August 2023, holding that human authorship is a fundamental requirement of US copyright law.

Thaler appealed, and the federal appeals court upheld the lower court's decision. In March 2026, the US Supreme Court declined to take up the case, letting the appeals court ruling stand. This effectively settled the question at the highest judicial level without a full Supreme Court opinion: under current US law, purely AI-generated works do not receive copyright protection because they lack a human author.

The practical consequence is significant. If you type a prompt into Midjourney, DALL-E, or any other generator, and the AI produces an image without further substantial human creative intervention, that image is not copyrightable. Anyone can use, reproduce, modify, or redistribute it without infringing copyright. You cannot sue someone for using an AI image you generated if the image itself has no copyright protection.

When AI-Assisted Images Can Be Copyrighted

The distinction between "AI-generated" and "AI-assisted" is crucial. The Copyright Office has been clear that works combining AI-generated elements with substantial human creative expression can receive protection, but only on the human-authored portions.

The landmark example came from the graphic novel "Zarya of the Dawn" by Kris Kashtanova. The Copyright Office granted copyright registration for the overall work, including Kashtanova's text and arrangement of images, but specifically excluded copyright protection for the individual AI-generated images within the novel. The human creative choices in selecting, arranging, and contextualizing the AI images were protectable, the AI-generated images themselves were not.

This principle creates a spectrum of protection based on the degree of human creative involvement. At one end, a single AI-generated image from a text prompt receives no protection. At the other end, a complex work where AI-generated elements are significantly modified, combined, and integrated into a larger human-directed creative composition can receive protection on the human-authored elements.

Activities that strengthen a copyright claim for AI-assisted works include substantially editing or modifying AI-generated images in a photo editor, compositing multiple AI-generated elements into a new arrangement, combining AI-generated imagery with human-created text, illustrations, or designs, and making creative decisions about selection and arrangement that reflect original authorship. The more human creative judgment involved in transforming AI output into a finished work, the stronger the copyright claim.

Activities that do not establish copyright include writing the text prompt itself (the Copyright Office has indicated that prompts alone are not sufficient creative input), selecting the "best" image from multiple AI generations, and making minor adjustments like cropping, color correction, or formatting that do not constitute original creative expression.

What This Means for Commercial Use

The lack of copyright protection for AI-generated images does not prevent you from using them commercially. Copyright determines who owns the exclusive rights to a creative work, preventing others from copying it. If an AI-generated image has no copyright, it means nobody owns the exclusive rights, including you. You can use it freely, but so can anyone else who obtains a copy.

This creates a practical business consideration. If you use an AI-generated image in your advertising, competitors could legally use the same image (if they access it) without paying you or asking permission. The lack of exclusivity is the real commercial impact, not a prohibition on use. For social media content, blog illustrations, and marketing materials where exclusivity is less critical, this is rarely a practical problem. For brand-defining imagery, logos, and assets you need exclusive control over, the lack of copyright protection is a meaningful limitation.

Platform terms of service provide an additional layer of usage rights separate from copyright law. Midjourney, DALL-E, Leonardo AI, and other generators grant users commercial usage licenses as part of their paid subscriptions. These licenses give you permission to use the generated images commercially under the platform's terms, regardless of the copyright question. What they do not give you is the ability to prevent others from using identical or similar AI-generated images.

The Training Data Copyright Question

A separate but related legal issue concerns whether AI image generators infringe on the copyrights of artists whose work was used in the training data. Several high-profile lawsuits, including cases filed against Stability AI, Midjourney, and DeviantArt by groups of artists, allege that scraping copyrighted images from the internet to train AI models constitutes unauthorized reproduction of those works.

These cases have not produced definitive rulings as of mid-2026. The central legal question is whether training an AI model on copyrighted data constitutes fair use, a doctrine that permits certain uses of copyrighted material without permission. The defendants argue that training is transformative (the AI learns patterns rather than copying images) and does not substitute for the original works. The plaintiffs argue that the AI can reproduce artistic styles learned from their copyrighted work, effectively competing with the original artists.

This question matters for businesses because an unfavorable ruling could theoretically affect the legal status of images generated by models trained on copyrighted data. In practice, even if courts find that training constitutes infringement, liability would most likely fall on the model developers (Stability AI, Midjourney, OpenAI) rather than individual users of the generated output. No court has suggested that end users face liability for using AI-generated images.

Adobe Firefly addresses this risk directly by training exclusively on licensed content, making it the only major generator that eliminates training data copyright concerns entirely. For businesses operating in risk-averse environments, Firefly's approach provides the strongest legal foundation for commercial AI image use.

Practical Guidelines for Businesses

For low-stakes commercial use (social media posts, blog illustrations, internal materials, email marketing), the copyright question is largely academic. Use AI-generated images freely under your platform's commercial license terms. The lack of copyright means competitors could theoretically use the same images, but in practice, this almost never happens because the same prompt produces different images each time.

For high-stakes commercial use (national advertising campaigns, product packaging, book covers, brand-defining imagery), take additional precautions. Use Adobe Firefly for maximum legal protection. Substantially modify AI-generated images with original human creative work to strengthen potential copyright claims. Consult legal counsel about your specific risk tolerance and use case. Consider whether the lack of copyright exclusivity is acceptable for your particular application.

For works you want to copyright, ensure substantial human creative input beyond the prompt. Combine AI-generated elements with human-created content. Document your creative process showing human authorship decisions. Register the work with the Copyright Office, disclosing the AI-generated elements, and claim copyright only on the human-authored portions. The Copyright Office specifically requires disclosure of AI-generated content in registration applications.

Keep records of your prompts, generation settings, platform used, and any modifications made. While this documentation does not create copyright protection by itself, it establishes a record of your creative process that supports any claims of human authorship in the modified work.

Where the Law Is Heading

Several legislative proposals in the US Congress would address AI-generated content copyright directly, though none have passed as of mid-2026. The most likely legislative outcome is a formal codification of the current Copyright Office position (human authorship required) with additional provisions for transparency and disclosure requirements.

The EU AI Act, which is being phased in through 2026, includes transparency requirements for AI-generated content in certain contexts but does not address copyright ownership directly. The EU's approach focuses on labeling and disclosure rather than granting or denying copyright protection to AI output.

International harmonization remains unlikely in the near term. Different countries will continue to take different approaches, creating a patchwork of rules that businesses operating globally need to navigate. For practical purposes, the US position (no copyright for purely AI-generated works) is the most influential and the one most businesses should plan around.