AI-Generated Art and the Copyright Crisis: What Courts Are Actually Deciding

The $100,000 Question

Can a machine create something that a human can own? That’s the question at the heart of the AI copyright crisis, and it’s no longer theoretical. In 2022, the US Copyright Office denied registration for “Théâtre D’opéra Spatial,” an AI-generated image created by Jason Allen using Midjourney, which had won the Colorado State Fair’s digital arts competition. In 2023, a federal judge ruled that Stephen Thaler’s AI system, the “Creativity Machine,” could not be listed as an inventor on a patent. In 2024, a court held that AI-generated works without sufficient human authorship receive no copyright protection at all. The legal framework is taking shape, case by case, and the implications for artists, technology companies, and the creative economy are enormous.

The stakes are hard to overstate. If AI-generated works can’t be copyrighted, then anyone can freely use them — which would gut the business models of the generative AI companies selling access to those outputs. But if AI-generated works can be copyrighted, then the human creators whose work was used to train the AI systems may have a claim to ownership of the outputs. Both paths lead to disruption. The courts are trying to thread a needle, and the thread keeps getting thinner.

What the Copyright Office and Courts Have Decided

The legal landscape is evolving rapidly, but some clear positions have emerged:

The human authorship requirement: In March 2023, the US Copyright Office issued guidance stating that works generated by AI “without any creative contribution from a human actor” are not copyrightable, because copyright law protects “the fruits of intellectual labor” that are “founded in the creative powers of the [human] mind.” The guidance established a middle path: works that combine human creativity with AI assistance may be copyrightable, but only the human-authored portions receive protection. A human who uses AI as a tool — the way a photographer uses a camera — may claim copyright over the result if their creative input is sufficient. A human who merely types a prompt and accepts the AI’s output may not.

The Thaler decisions: Stephen Thaler’s attempts to register AI-generated works (and to list his AI system as a patent inventor) have been rejected at every level. The US Patent and Trademark Office and the Copyright Office both held that non-human entities cannot be authors or inventors. A federal district court affirmed this in 2023, and the decision was upheld on appeal. The principle is now well-established in US law: only humans can be authors.

The “Naruto” monkey precedent: The courts have consistently applied the logic of the famous “monkey selfie” case — in which a macaque named Naruto was held to lack standing to sue for copyright infringement of a photo it took — to AI systems. If a monkey can’t own a copyright, neither can a machine. The reasoning is that copyright law’s purpose is to incentivize human creativity, and non-human creators have no need for incentives.

The training data lawsuits: A separate but related battle concerns the use of copyrighted works to train AI systems. Artists including Sarah Andersen, Kelly McKernan, and Karla Ortiz have sued Stability AI, Midjourney, and DeviantArt, alleging that their copyrighted works were used without permission to train image-generation models. Getty Images has sued Stability AI for using its stock photo library. The Authors Guild has sued OpenAI and Microsoft over the use of copyrighted books in training data. These cases, still winding through courts, will determine whether “training on copyrighted material” constitutes fair use or infringement — a question with billions of dollars in implications.

The Fair Use Battlefield

The central legal question in the training data lawsuits is whether using copyrighted works to train AI models constitutes fair use. The AI companies argue that training is transformative — the model doesn’t reproduce the copyrighted works; it learns statistical patterns from them, the way a human artist learns by studying other artists’ work. The plaintiffs argue that training on copyrighted material without permission or compensation is fundamentally different from human learning, because the AI system can ingest millions of works at a scale no human could, and because the outputs can compete directly with the works they were trained on.

The fair use analysis considers four factors: the purpose of the use, the nature of the copyrighted work, the amount used, and the effect on the market. The AI companies have strong arguments on the first factor (training is arguably transformative) but face challenges on the fourth (AI-generated art can substitute for human-created art, harming the market for the original works). The courts have yet to issue a definitive ruling, and the outcome will likely vary by case, by model, and by the specific copyrighted works involved.

The economic stakes are enormous. If courts rule that training on copyrighted material requires a license, the generative AI industry would face a wave of licensing obligations — potentially retroactive — that could fundamentally reshape its economics. If courts rule that training is fair use, the human creators whose work was ingested would receive no compensation for what may be the most valuable use of their work in history. Either outcome will produce winners and losers, and the transition will be messy.

What This Means for Creators

For working artists, the AI copyright crisis is both a threat and an opportunity. The threat is obvious: AI-generated art can now be produced in seconds, at near-zero cost, and in styles that mimic (some would say plagiarize) specific artists. The opportunity is less obvious but real: artists who learn to use AI as a tool — incorporating it into their workflow, using it for ideation, iteration, and production — can dramatically increase their output and explore creative directions that weren’t practical before.

The key legal distinction for creators is the level of human authorship. An artist who uses AI to generate a base image and then substantially modifies it — compositing, editing, adding original elements — has a stronger copyright claim than an artist who types a prompt and posts the raw output. The Copyright Office has signaled that “sufficient human authorship” is the threshold, and the cases decided so far suggest that meaningful creative control and modification matter. Artists who want copyright protection for their AI-assisted work should document their creative process and emphasize the human contributions.

The deeper question — whether AI art is “real” art — is not one the courts can answer. What they can answer, and are answering, is who owns what. The answers are coming, case by case, and they’re reshaping the creative economy in real time. For creators, the message is clear: adapt or risk being left without legal protection for your work.

The International Dimension

The AI copyright crisis isn’t just an American legal drama. The European Union’s AI Act, which entered into force in 2024, includes transparency requirements for AI training data — companies must disclose what copyrighted material was used to train their models. The UK’s Intellectual Property Office is conducting a consultation on AI and copyright that could establish a licensing framework. Japan has taken a permissive approach, explicitly allowing AI training on copyrighted material under certain conditions. The result is a patchwork of national approaches that creates enormous compliance complexity for AI companies that operate globally.

The most consequential battle may be between the United States and the European Union. If the EU requires AI companies to license training data while the US does not, the competitive landscape could shift dramatically — AI development might gravitate toward the jurisdiction with the most favorable legal environment. Alternatively, if major markets converge on a licensing requirement, the entire generative AI industry’s cost structure would change overnight. The outcome is impossible to predict from current case law, but the direction of travel is toward more regulation, not less, and the companies that prepare for it will fare better than those that resist.