Authorship in the Age of AI: Copyright and Legal Challenges Across Jurisdictions

Introduction 

Generative artificial intelligence challenges a core assumption of copyright law: that creative works are the product of human intellect. AI can now autonomously produce images, music, texts, designs, and legal documents that often resemble human work, yet its mode of creation conflicts with copyright’s requirements of authorship, originality, and intention.

Copyright protects original expression fixed in a tangible medium, rewarding human skill, labour, and judgment, while excluding ideas, facts, and mechanical reproductions. It assumes a human author who makes creative choices and bears responsibility. Generative AI disrupts this model by producing outputs through pattern recognition rather than conscious decision-making, blurring the line between authorship and automation.

This raises three key questions: can AI be recognised as an author; if not, who owns AI-generated works; and does training AI on copyrighted material constitute infringement? Litigation and regulation in the UK, EU, and US show contrasting approaches, reflecting different philosophies on innovation, enforcement, and creator protection. This article examines whether copyright law can adapt to generative AI without undermining human creativity or stifling innovation.

United Kingdom: Reactive Human-Centred Framework

UK copyright law provides a limited framework for non-human creation. Section 9(3) of the Copyright, Designs and Patents Act 1988 states that the author of a computer-generated work is the person who makes the arrangements necessary for its creation1. Originally designed for deterministic software, it assumed that humans retained meaningful control over the outcome.

Courts have consistently interpreted originality through a human-centred lens. Following Infopaq International A/S v Danske Dagblades Forening2, originality requires the author’s own intellectual creation expressed through skill, labour, and judgment. Domestic cases, including Temple Island Collections Ltd v New English Teas Ltd3, confirm that technical effort alone is insufficient. Modern generative AI exposes a tension: outputs created without human decision-making may struggle to meet the originality threshold.

Getty Images (US) Inc v Stability AI Ltd4 represents the first major UK case confronting this tension. Getty alleged that Stability AI scraped millions of copyrighted images to train its Stable Diffusion model without permission, bringing claims of copyright, database right, trademark infringement, and passing off. Many claims were abandoned because the training occurred offshore, highlighting the territorial limits of UK copyright law.

Mrs Justice Joanna Smith rejected secondary copyright infringement claims, holding that Stable Diffusion did not store or reproduce copies. Instead, it encoded statistical relationships, which do not constitute copying under sections 22 and 23 of the CDPA5. Limited trademark liability arose where outputs resembled Getty watermarks, but copyright liability was not established6.

The case underscores structural weaknesses in the UK approach. Territoriality enables regulatory arbitrage through offshore training, the lack of dataset transparency leaves rights holders at a disadvantage, and whether AI-generated outputs can ever qualify as original remains unresolved. UK law therefore remains reactive, relying on case specific adjudication using concepts developed for human authorship.

European Union: Proactive Safeguards and Licensing Controls

The European Union adopts a more proactive approach. Existing directives, like the Information Society Directive7, continue to protect original works, while the Digital Single Market (DSM) Directive8 introduces specific rules for text and data mining. Under Article 4 of the DSM Directive, commercial mining is only allowed if rights holders have not opted out, preserving licensing markets and placing responsibility on AI developers. This reflects the principle that large-scale data use should remain under author control. Unlike the UK, EU law builds copyright considerations directly into AI development.

The EU Artificial Intelligence Act 20249 reinforces this approach. Although not a copyright law, it requires developers of high-risk AI systems to maintain transparency, document their datasets, and manage risks. These measures help rights holders trace potential infringements and improve enforcement.

EU copyright remains explicitly human-centred: originality depends on the author’s own intellectual creation, so purely AI-generated works are excluded. While this denies AI proprietary rights, it strengthens human creators’ position during the training process. Overall, the EU prioritises preventive safeguards and licensing controls over resolving disputes after the fact, even if compliance adds extra burdens for developers.

United States: Human Authorship with Flexible Fair Use

United States copyright law similarly insists on human authorship. Guidance issued by the US Copyright Office in 202310 confirms that protection extends only to works involving human creative input, while material generated autonomously by AI systems is not registrable. Where a human meaningfully selects, edits or arranges AI-generated material, the resulting work may satisfy the minimal creativity threshold, but standalone AI outputs generally fall outside protection.

US copyright law also operates within a flexible fair use framework, which may shield certain training activities, particularly where outputs do not reproduce protected expression. This approach prioritises innovation while maintaining a clear doctrinal boundary between human authorship and automated generation.

At the state level, California has emerged as the most significant US regulator of artificial intelligence. By passing SB 5311, California became the first state to directly regulate developers of frontier foundation models, defined by computational scale rather than downstream use. Unlike statutes such as the Colorado AI Act12, which adopt a risk-based approach focused on high-risk applications, SB 53 targets the development stage of large-scale models. Although not a copyright statute, its transparency, documentation and disclosure requirements create evidentiary trails that may assist rights holders in identifying unauthorised dataset use. Given California’s concentration of both AI developers and creative industries, this regime is likely to exert de facto national influence.

Conclusion 

Generative AI forces us to rethink not just who creates, but what creation means in a digital age. Beyond legal technicalities, it exposes a tension between human ingenuity and algorithmic capability. The law must recognise that value lies as much in the choices, context, and curation that humans provide as in the outputs themselves. Any effective framework will require more than enforcement; it demands a cultural shift toward transparency, accountability, and collaboration between creators and technologists. By integrating mechanisms that make datasets traceable and use auditable, copyright law can evolve from a reactive system into a proactive one, shaping the development of AI while safeguarding both the rights and livelihoods of human creators.

Footnotes 

1 Copyright, Designs and Patents Act 1988, s 9(3).
2 Infopaq International A/S v Danske Dagblades Forening (C-5/08) [2009] ECR I-6569.
3 Temple Island Collections Ltd v New English Teas Ltd [2012] EWPCC 1.
4 Getty Images (US) Inc v Stability AI Ltd [2025] EWHC 2863 (Ch).
5 Copyright, Designs and Patents Act 1988, ss 22–23.
6 Trade Marks Act 1994, s 10.
7 Directive 2001/29/EC of the European Parliament and of the Council on the Harmonisation of Certain Aspects of Copyright and Related Rights in the Information Society [2001] OJ L167/10.
8 Directive (EU) 2019/790 on Copyright and Related Rights in the Digital Single Market [2019] OJ L130/92.
9 European Union, Artificial Intelligence Act 2024.
10 US Copyright Office, ‘Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence’ (16 March 2023) 88 Fed Reg 16190.
11 California Legislature, Transparency in Frontier Artificial Intelligence Act, SB 53 (2025).
12 Colorado General Assembly, Colorado AI Act (2023).

Bibliography

Primary Legislation and Directives
Copyright, Designs and Patents Act 1988 (c.48)
Trade Marks Act 1994 (c.26)
Directive 2001/29/EC of the European Parliament and of the Council of 22 May 2001 on the Harmonisation of Certain Aspects of Copyright and Related Rights in the Information Society [2001] OJ L167/10
Directive (EU) 2019/790 of the European Parliament and of the Council of 17 April 2019 on Copyright and Related Rights in the Digital Single Market [2019] OJ L130/92
European Union, Artificial Intelligence Act 2024

Cases
Infopaq International A/S v Danske Dagblades Forening (C-5/08) [2009] ECR I-6569
Temple Island Collections Ltd v New English Teas Ltd [2012] EWPCC 1
Getty Images (US) Inc and others v Stability AI Ltd [2025] EWHC 2863 (Ch)

Official Guidance
US Copyright Office, ‘Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence’ (Federal Register, 16 March 2023) 88 Fed Reg 16190

Commentary
Dentons, ‘AI gains stability: UK court finds no secondary copyright infringement in Getty Images v Stability AI’ (5 November 2025) https://www.dentons.com/en/insights/articles/2025/november/5/ai-gains-stability-uk-court-finds-no-secondary accessed 8 December 2025
Ward Hadaway, ‘Getty Images v Stability AI – High Court Judgment’ (2025) https://www.wardhadaway.com/insights/updates/getty-images-v-stability-ai-high-court-judgment/ accessed 8 December 2025
Latham & Watkins LLP, ‘California Assumes Role as Lead US Regulator of AI’ (15 October 2025) https://www.lw.com/en/insights/california-assumes-role-as-lead-us-regulator-of-aiaccessed 8 December 2026
Reuters, ‘Getty Images largely loses landmark UK lawsuit over AI image generator’ Reuters (4 November 2025) https://www.reuters.com/sustainability/boards-policy-regulation/getty-images-largely-loses-landmark-uk-lawsuit-over-ai-image-generator-2025-11-04/ accessed 8 December 2026

Image Credits: Growtika on Unsplash