ClearFactAll topics

Should AI-generated content creators have legal copyright ownership of their outputs?

5 peer-reviewed studies for, 5 against

Key takeaways

Stronger evidence: Against

Verdict

The evidence weighs against giving people copyright ownership of AI-generated outputs as such, while supporting protection for the genuinely creative human contributions behind them. On the "against" side, Lemley (Science and Technology Law Review, 2024) argues that the AI does the bulk of the expressive work copyright exists to reward, so that work will not be protected even when the prompts are creative. Gervais (GRUR International, 2020) argues that once a system crosses an autonomy threshold and makes the expressive choices itself, there is no human author to protect. Craig and Kerr (Robot Law, 2025) call AI authorship a category mistake, and Chen (Journal of Intellectual Property Law & Practice, 2025) warns that protecting AI output regardless of human input creates an originality paradox and a "contributor thicket" whose harms may outweigh the benefits. Mezei (UFITA, 2020) argues such output belongs in the public domain. On the "for" side, the strongest point is that human-directed AI work can already qualify: Hugenholtz and Quintais (IIC, 2021) show EU law is flexible enough to protect AI-assisted output when a person makes the creative choices, and a 2023 analysis of the UK's section 9(3) argues the user should own outputs when their instructions are themselves original. China's Beijing Internet Court has recognized an AI user as the author of an AI-generated image (Wen, 2024), and Sun and Zhou (Journal of World Intellectual Property, 2026) propose a human-AI collaborative authorship model. Abbott and Rothman (2024) add a utilitarian argument that protection would encourage more creative output. Notably, nearly every "for" source ties ownership to sufficient human contribution rather than to the output itself. The stronger position is therefore against blanket ownership of AI outputs: protection should follow demonstrable human creativity in selecting, arranging or directing the work, which is close to the current U.S. approach, while the incentive argument for protecting machine-made expression remains largely theoretical.

Strongest study on each side

Ranked by citation count.

Strongest study for

Copyright and Artificial Creation: Does EU Copyright Law Protect AI-Assisted Output? (IIC - International Review of Intellectual Property and Competition Law, 2021)

145 citations

Hugenholtz and Quintais carry out a doctrinal analysis of "work", "originality", "creative freedom" and authorship in EU copyright law and Court of Justice case law. They develop a four-step test for when AI-assisted output qualifies as an original work, and conclude that current EU rules are generally suitable and flexible enough to protect AI-assisted output and assign authorship to the human who made the creative choices.

Strongest study against

How Generative AI Turns Copyright Law Upside Down (Science and Technology Law Review, 2024)

82 citations

Lemley argues that with generative AI, creativity increasingly lies in asking the right questions rather than creating the answers. Writing prompts may sometimes be creative, but the AI does the bulk of the work copyright traditionally rewards, and that work will not be protected. The result inverts what copyright prizes and strains the idea-expression dichotomy and the substantial-similarity test.

Open the interactive page →

The evidence

For (5)

Copyright and Artificial Creation: Does EU Copyright Law Protect AI-Assisted Output? (IIC - International Review of Intellectual Property and Competition Law, 2021)

145 citations

Hugenholtz and Quintais carry out a doctrinal analysis of "work", "originality", "creative freedom" and authorship in EU copyright law and Court of Justice case law. They develop a four-step test for when AI-assisted output qualifies as an original work, and conclude that current EU rules are generally suitable and flexible enough to protect AI-assisted output and assign authorship to the human who made the creative choices.

Dealing with AI-generated works: lessons from the CDPA section 9(3) (Journal of Intellectual Property Law & Practice, 2023)

27 citations

Examines the UK's section 9(3) rule on computer-generated works and how users' prompts affect whether copyright subsists in AI-generated outputs. Supports the CDPA's human-centred approach and argues that AI-generated works should belong to the user who gives the instructions, as long as those instructions are themselves sufficiently original, and that this should be written expressly into the statute.

How much human contribution is needed for "ownership" of AI-generated content: A comparison of copyright determination for generative AI in China and the United States (The Journal of World Intellectual Property, 2026)

3 citations

Sun and Zhou compare key cases in which Chinese courts affirmed copyright ownership for AI users while the U.S. Copyright Office declined registration. They trace the divergence to how each system weighs human contribution in the generation process, and propose a human-AI collaborative authorship model under which users who make meaningful contributions can own the resulting content.

Against (5)

How Generative AI Turns Copyright Law Upside Down (Science and Technology Law Review, 2024)

82 citations

Lemley argues that with generative AI, creativity increasingly lies in asking the right questions rather than creating the answers. Writing prompts may sometimes be creative, but the AI does the bulk of the work copyright traditionally rewards, and that work will not be protected. The result inverts what copyright prizes and strains the idea-expression dichotomy and the substantial-similarity test.

The death of the AI author (Robot Law: Volume II, 2025)

9 citations

Craig and Kerr argue that the threshold for authorship does not depend on how sophisticated AI becomes: the very idea of AI authorship is a category mistake resting on a romanticised view of both authorship and AI. They propose a relational account of authorship that reorients debates about whether copyright should subsist in AI-generated works at all.

Reviving 'computer-generated works': should Hong Kong copyright law adapt the rule to harness AI opportunities? (Journal of Intellectual Property Law & Practice, 2025)

1 citations

Chen challenges the Hong Kong government's view that existing computer-generated-works provisions already protect AI-generated content. Applying them creates an "originality paradox" and a "contributor thicket" in which too many people could claim authorship, and protecting AI output regardless of human input undermines the very policy goals it is meant to serve, with harms likely outweighing benefits.

The debate

Maya: If I spend hours crafting prompts, choosing between hundreds of generations and editing the result, I should own it. China's Beijing Internet Court agreed: it recognized the AI user as the author of an AI-generated image, as Wen reported in 2024.

Daniel: But look at why the court found authorship: the user's own choices. That's the human contribution, not the output. Lemley's point in the Science and Technology Law Review is that with generative AI, the machine does the bulk of the expressive work, and that's exactly the part copyright has never protected without a human author.

Maya: Hugenholtz and Quintais showed in 2021 that EU law already handles this. Their four-step test asks whether a person made free and creative choices during conception, execution and redaction. If so, AI-assisted output is a protected work. The law doesn't need to change at all.

Daniel: I agree with that, and notice what it concedes: protection depends on the human's creative choices, not on the fact that something was generated. Gervais makes the same point from the other direction. Once the system crosses an autonomy threshold and makes the expressive choices itself, there's no author left to protect.

Maya: Then think about incentives. Abbott and Rothman argue from copyright's own utilitarian purpose: if AI outputs fall into the public domain, businesses have less reason to invest in creating and distributing them.

Daniel: That incentive argument is mostly theoretical. People are already generating enormous amounts of AI content with no copyright at all. And Chen's 2025 analysis of Hong Kong's proposal shows the costs: protecting output regardless of human input creates an originality paradox and a thicket of people who could all claim to be authors.

Maya: So what about the UK's computer-generated works rule? It already gives rights to the person who made the arrangements.

Daniel: Even there, the 2023 analysis of section 9(3) argues the user should own AI output only when their instructions are themselves original. And Craig and Kerr go further: calling the AI an author is a category mistake. Authorship is a human, relational idea.

Maya: Then we agree on more than it seems. People who genuinely direct, select and shape AI work should be protected for that creativity.

Daniel: Yes. Protect the human creativity, not the raw output. Anything a machine produces on its own should stay in the public domain, which is where Mezei argues it belongs.

Frequently asked questions

Can you copyright something made with AI?

It depends on the human contribution. The evidence suggests protection can follow demonstrable human creativity in selecting, arranging or directing the work, but not the AI-generated expression itself.

Are AI prompts enough to claim copyright?

Not necessarily. One analysis argues that even creative prompts leave the bulk of the expressive work to the AI, so that work is not protected. A UK-focused analysis argues the user should own outputs when their instructions are themselves sufficiently original.

How do the U.S. and China differ on AI-generated works?

Chinese courts have affirmed copyright ownership for AI users in key cases, while the U.S. Copyright Office has declined registration. Researchers trace the difference to how each system weighs human contribution in the generation process.

Would copyright for AI outputs encourage more creativity?

A utilitarian argument holds that protection would give people and businesses an incentive to produce and share more creative output. However, the incentive argument for protecting machine-made expression remains largely theoretical, and critics warn of an originality paradox and too many competing authorship claims.

Related questions