AI-generated images created with tools like Midjourney cannot be copyrighted in the United States, and the legal status of training these systems on copyrighted artwork remains deeply uncertain. This fundamental copyright gap is now at the center of a legal battle that could reshape how generative AI companies operate and what protections creators actually have.

Can AI Art Actually Be Copyrighted?

The short answer is no, at least not in the way most creators expect. The U.S. Copyright Office has long held that works created solely by artificial intelligence, even if prompted by a human, cannot be copyrighted. This position was tested directly in 2022 when Kristina Kashtanova sought copyright protection for "Zarya of the Dawn," an 18-page graphic novel created entirely with Midjourney.

Initially, the Copyright Office granted the registration. But just months later, it reversed course and partially canceled the copyright, explaining that the images themselves were "not the product of human authorship" but rather unpredictable outputs generated by the AI system based on text prompts. The office deemed Kashtanova's editing of the images "too minor and imperceptible to supply the necessary creativity for copyright protection." The book's text and overall arrangement remained protected, but the visual elements, which were the core of the work, did not.

What Counts as Human Authorship in AI Collaboration?

The line between human creativity and machine generation is blurry, and it matters legally. According to intellectual property experts, the threshold depends on how much control a human author exercises over the AI's output. Daniel Gervais, a professor at Vanderbilt Law School who specializes in intellectual property law, explained the distinction.

"If a machine and a human work together, but you can separate what each of them has done, then copyright will only focus on the human part. If the human and machine's contributions are more intertwined, a work's eligibility for copyright depends on how much control or influence the human author had on the machine's outputs. It really needs to be an authorial kind of contribution. In that case, the fact that you worked with a machine would not exclude copyright protection," said Daniel Gervais.

Daniel Gervais, Professor of Intellectual Property Law at Vanderbilt Law School

In practice, this means that simply typing a prompt into Midjourney and accepting the generated output likely does not qualify as sufficient human authorship. The Copyright Office's policy, released after the "Zarya of the Dawn" decision, reinforces this stance: if a human types a prompt and the machine generates complex visual works in response, the "traditional elements of authorship" have been executed by AI, not a human.

How to Protect Your Creative Work When Using AI Tools

For creators working with AI image generators, understanding the legal landscape is essential. Here are the key considerations:

• Significant Human Editing: The more substantial and creative your edits to AI-generated images, the stronger your copyright claim. Minor adjustments will not meet the threshold for copyright protection.
• Original Conceptual Work: Focus on creating original prompts, compositions, and artistic direction that demonstrate your unique creative vision, rather than relying on the AI to generate the entire concept.
• Documentation of Process: Keep detailed records of your creative decisions, iterations, and the reasoning behind your choices when working with AI tools, as this can support claims of human authorship.

The Training Data Lawsuit Crisis

While the copyright status of AI-generated images remains murky, an even larger legal battle is unfolding over how these systems are trained. Multiple lawsuits are challenging whether companies like Stability AI and OpenAI can legally use copyrighted works to train their models without permission or compensation.

Getty Images sued Stability AI, the company behind Stable Diffusion, for copying and processing millions of copyrighted images and their associated metadata without permission. The New York Times filed lawsuits against both OpenAI and Microsoft in 2023, and again against Perplexity in 2025, claiming these companies used millions of Times articles to train their AI models without compensation. Additional newspapers, including The New York Daily News and the Chicago Tribune, have also sued Microsoft and OpenAI.

The stakes are enormous. If courts find that these companies illegally used copyrighted content to train their models, they could be forced to destroy their training datasets and rebuild them from scratch, according to NPR's reporting on the New York Times case.

The Fair Use Question That Could Change Everything

Currently, the fair use doctrine permits the use of copyrighted material under certain conditions without needing the copyright holder's permission. This legal principle has historically allowed limited use of copyrighted works for purposes like criticism, commentary, and education. But generative AI companies are pushing this doctrine to its limits.

The question at the heart of these lawsuits is whether training an AI system on copyrighted works, especially when the outputs compete with the original works, qualifies as fair use. Creators and publishers argue that it does not, since the AI outputs directly compete with their original content. The courts have not yet definitively answered this question, but the pending cases could reshape copyright law for the entire AI industry.

For Midjourney users and other AI art creators, the implications are significant. If courts rule against fair use for AI training, the datasets used to train these systems may need to be substantially altered or rebuilt. This could affect the quality and capabilities of future versions of these tools. More immediately, creators need to understand that the images they generate with Midjourney cannot be copyrighted as standalone works, and the legal status of the training data behind the tool remains contested.