Section 9(3) of the UK Copyright, Designs and Patents Act 1988 (CDPA) deems the author of a computer-generated work to be “the person by whom the arrangements necessary for the creation of the work are undertaken”. Atilla Söğüt’s article in the Journal of Intellectual Property Law & Practice (2024) proposed that, for generative-AI outputs, this person should be the prompting user — but only when the user’s directions to create are themselves sufficiently original — while acknowledging that no existing case law fixes the threshold at which a short prompt becomes an original “author’s own intellectual creation”, that the interaction of short-text precedents with iterative prompting is unexplored, and that the pastiche exception and the idea/expression dichotomy are undeveloped for prompts that direct AI to mimic specific artists. The position remains unsettled in the existing authorities and scholarship: US Copyright Office guidance denies prompt-only authorship without fixing any threshold; Chinese first-instance courts are split (the Beijing Internet Court’s 2023 prompt-based protection versus a 2025 Suzhou denial); the first English generative-AI trial judgment, Getty Images v Stability AI [2025] EWHC 2863 (Ch), left prompts and pastiche untouched; and scholarship through 2026 still proposes competing frameworks rather than settled standards. The problem — to articulate a determinate, doctrinally grounded standard for when prompting grounds authorship of AI-generated works, and for how style-mimicking prompts are treated under the idea/expression dichotomy and the pastiche exception — remains open.

Background

UK copyright law contains a provision written before generative models existed. Section 9(3) of the Copyright, Designs and Patents Act 1988 (CDPA) deals with “computer-generated works” — works “generated by computer in circumstances such that there is no human author” — and deems their author to be “the person by whom the arrangements necessary for the creation of the work are undertaken”, with a shortened term of 50 years from making. A user of a generative-AI system neither paints nor writes the output; they issue textual instructions (a “prompt”) and may then refine, re-run, adjust parameters, and select among candidate outputs. Whether and when that activity makes the user the “person by whom the arrangements necessary” are undertaken — and whether it supplies the originality copyright requires — is contested.

Originality, in the UK and EU standard applied by the Court of Justice of the European Union (the author’s own intellectual creation, reflecting the author’s free and creative choices, as summarized in Ampovska 2025), is qualitative. The source article’s cited authorities illustrate the short-text boundary: the UK decision in Exxon (Exxon Corp v Exxon Insurance Consultants International Ltd) refused copyright in a single coined word, and the article pairs it with a US decision (“Stern v. Does”) said to refuse protection to a single sentence, while recognizing that no absolute minimum length governs protection. A one-off two-word prompt and a multi-clause specification plainly sit differently, but no criterion in the case law says where the line falls — and prompting is usually iterative, so creative choices may be distributed across a chain of individually unremarkable instructions and selections rather than present in any single prompt.

On the infringement side, when a prompt directs the AI to imitate a named artist or style, two doctrines control the risk. The idea/expression dichotomy holds that ideas, methods and styles in the abstract are free for all to use and only particular expression is protected — the US Copyright Office has characterized prompts as “instructions that convey unprotectible ideas” (Copyright and Artificial Intelligence, Part 2, 2025). The UK fair-dealing exception for parody, caricature and pastiche (CDPA s 30A, added in 2014) permits certain transformative reuses; a “pastiche” is a work that imitates another’s style or mixes recognizable elements from multiple works, and the exception has almost no case law even outside the AI context.

Several results frame but do not answer the question. Stephen Thaler’s attempts to register works authored solely by his machines failed: the English courts refused AI inventorship for the DABUS system (a device Thaler presented as the “Device for the Autonomous Bootstrapping of Unified Sentience”) in the patent litigation ([2020] EWHC 2412 (Pat), affirmed [2021] EWCA Civ 1374), and the D.C. Circuit in Thaler v. Perlmutter (2025) required a human author for the image “A Recent Entrance to Paradise”, said to be generated by his “Creativity Machine”; English rulings on his parallel copyright claim are reported to the same effect. In all of these proceedings Thaler deliberately disclaimed human creative control, so none decides when a human’s prompting suffices. The US Copyright Office concluded that, with current technology, prompts alone do not provide sufficient control to make users authors, while expressly leaving every assessment case-by-case. In China, the Beijing Internet Court protected a Stable Diffusion image because the user’s prompt design, keyword selection and iterative parameter tuning constituted human intellectual input (Li v. Liu, November 2023), whereas the Suzhou Intermediate People’s Court in 2025 denied copyrightability of AI-generated images for the first time. The first English generative-AI trial judgment, Getty Images (US) Inc & Ors v Stability AI Limited [2025] EWHC 2863 (Ch), ruled on training, secondary copyright infringement and trade mark but says nothing about prompt originality or pastiche. It is against this landscape that the source article’s gateway proposal — copyright in the output conditional on the originality of the user’s directions to create — leaves open the threshold question this problem poses.

Problem Statement

Under UK copyright law (CDPA s 9(3), read with the “author’s own intellectual creation” originality standard shared with EU law), at what point does a generative-AI user’s prompting activity become sufficiently original to ground authorship of — or copyright subsistence in — the resulting computer-generated work? The source article’s proposal makes this the gateway question: the user should own the output only when the user’s directions to create are themselves sufficiently original. The criterion sought must be doctrinally derived and determinate, and must decide:

(1) whether a prompt of a given character, detail and length — from a two-word instruction to a multi-clause specification — is an “author’s own intellectual creation”, consistently with authorities that deny protection to very short texts while recognizing that no absolute minimum length governs;

(2) how the criterion treats iterative prompting — chains of individually short or unoriginal prompts, refinements, parameter adjustments and output selections whose aggregate may be more creative than any single prompt — including whether originality is assessed per prompt, across the whole prompting course, or in the selection and arrangement of the resulting outputs; and

(3) when a prompt directs the AI to imitate a specific artist or style, how the idea/expression dichotomy and the pastiche fair-dealing exception allocate the resulting risk of infringement between unprotected style and protected expression.

The single objective uniting these strands is a coherent account of the prompt’s legal function in generative-AI copyright — on the subsistence side (does the prompting make an author?) and the infringement side (does a mimicking prompt make an infringer?) — stated precisely enough that a court or copyright registry could apply it to concrete cases, and shown to reconcile or properly distinguish the existing UK, EU, US and Chinese authorities.

References

  1. Söğüt, Atilla (2024), “Dealing with AI-generated works: lessons from the CDPA section 9(3)”, Journal of Intellectual Property Law & Practice 19(1):43–54, DOI 10.1093/jiplp/jpad102, https://doi.org/10.1093/jiplp/jpad102
  2. U.S. Copyright Office (2025), “Copyright and Artificial Intelligence, Part 2: Copyrightability”, Report of the Register of Copyrights, January 2025, https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf
  3. Thaler v. Perlmutter, No. 23-5233 (United States Court of Appeals for the District of Columbia Circuit, decided 18 March 2025), https://www.copyright.gov/ai/docs/court-of-appeals-decision-affirming-refusal-of-registration.pdf
  4. Getty Images (US) Inc & Ors v Stability AI Limited [2025] EWHC 2863 (Ch) (High Court of England and Wales, 2025), https://caselaw.nationalarchives.gov.uk/ewhc/ch/2025/2863
  5. Thaler v The Comptroller-General of Patents, Designs and Trade Marks [2020] EWHC 2412 (Pat) (High Court of England and Wales, 2020), https://caselaw.nationalarchives.gov.uk/ewhc/pat/2020/2412
  6. Ampovska, Marija (2025), “Thresholds for Authorship and Originality in AI-generated and AI-assisted works”, Bratislava Law Review 9(2):91–108, DOI 10.46282/blr.2025.9.2.1136, https://doi.org/10.46282/blr.2025.9.2.1136
  7. He, Tianxiang (2024), “AI Originality Revisited: Can We Prompt Copyright over AI-Generated Pictures?”, GRUR International 73(4):299–307, DOI 10.1093/grurint/ikae024, https://doi.org/10.1093/grurint/ikae024
  8. Mazzi, Francesca (2024), “Authorship in artificial intelligence-generated works: Exploring originality in text prompts and artificial intelligence outputs through philosophical foundations of copyright and collage protection”, The Journal of World Intellectual Property 27(3):410–427, DOI 10.1111/jwip.12310, https://doi.org/10.1111/jwip.12310
  9. Fritz, Johannes (2025), “Understanding authorship in Artificial Intelligence-assisted works”, Journal of Intellectual Property Law & Practice 20(5):354–364, DOI 10.1093/jiplp/jpae119, https://doi.org/10.1093/jiplp/jpae119
  10. Bektaş, Ulaş Göker (2026), “The Algorithmic Authorship Threshold: identifying human contribution in AI-assisted creativity”, Journal of Intellectual Property Law & Practice 21(9):610–618, DOI 10.1093/jiplp/jpag078, https://doi.org/10.1093/jiplp/jpag078
  11. Tang, Han (2026), “Chinese court denied the copyrightability of AI-generated images for the first time in (2025) Su 0582 Min Zhong No. 4840”, Journal of Intellectual Property Law & Practice 21(3):139–142, DOI 10.1093/jiplp/jpag001, https://doi.org/10.1093/jiplp/jpag001
  12. Chen, Yang (2025), “Reviving ‘computer-generated works’: should Hong Kong copyright law adapt the rule to harness AI opportunities?”, Journal of Intellectual Property Law and Practice 20(9):584–593, DOI 10.1093/jiplp/jpaf035, https://doi.org/10.1093/jiplp/jpaf035
  13. Hall, James (2026), “Why Getty Images v Stability AI got secondary copyright infringement wrong”, Journal of Intellectual Property Law & Practice 21(8):538–545, DOI 10.1093/jiplp/jpag074, https://doi.org/10.1093/jiplp/jpag074
  14. Verma, Surbhi (2026), “AI Training, Copyright, and the Limits of Territorial Jurisdiction: A Case Analysis of Getty Images v Stability AI [2025] EWHC 2863 (Ch)”, SSRN preprint, DOI 10.2139/ssrn.6974038, https://doi.org/10.2139/ssrn.6974038
  15. kim, June (2026), “Who Authored the Spec? Evidence for the Line-drawing Question Thaler Left Open”, SSRN preprint, DOI 10.2139/ssrn.7140418, https://doi.org/10.2139/ssrn.7140418
  16. Bishop, Lea (2026), “Brief for Lea Bishop as Amicus Curiae Supporting Neither Party, Thaler v. Perlmutter, No. 25-449 (U.S., filed 26 January 2026)”, SSRN preprint, DOI 10.2139/ssrn.6159510, https://doi.org/10.2139/ssrn.6159510
  17. Sobel, Benjamin (2024), “Elements of Style: Copyright, Similarity, and Generative AI”, SSRN preprint, DOI 10.2139/ssrn.4832872, https://doi.org/10.2139/ssrn.4832872
  18. Izyumenko, Elena (2026), “Copyrighting Style? Reframing Style Imitation as Non-Literal Copying in EU Copyright Law: Implications for Generative AI and Fashion Dupes”, SSRN preprint, DOI 10.2139/ssrn.7206858, https://doi.org/10.2139/ssrn.7206858
  19. Graves, Franklin (2022), “Thaler Pursues Copyright Challenge Over Denial of AI-Generated Work Registration”, SSRN preprint, DOI 10.2139/ssrn.4192155, https://doi.org/10.2139/ssrn.4192155
  20. Zhao, Sichen (2025), “Determining the Originality of AI-Generated Works: A Prompt-Based Perspective”, Beijing Law Review 16(3):1390–1404, DOI 10.4236/blr.2025.163070, https://doi.org/10.4236/blr.2025.163070
  21. Jütte, Bernd Justin; Ryszkowska, Paulina (2026), “European Parliament and House of Lords publish reports on AI and copyright: policy directions for artificial intelligence and copyright in the EU and the UK”, Journal of Intellectual Property Law & Practice 21(9):654–659, DOI 10.1093/jiplp/jpag061, https://doi.org/10.1093/jiplp/jpag061