Can AI-generated content be protected by copyright? The state of play in the UK, EU and China

Generative AI models have impressed the world with their ability to generate text, image, audio and video content, of increasingly high quality and at increasingly rapid speed. This has been an exciting development in industries such as media, film and TV production, video games and publishing, where AI can be used to shortcut routine content production tasks that would otherwise be laborious and time consuming. Indeed, one of the most common use cases we have observed in recent years is use of AI to generate advertising and marketing content, vastly expanding the possibilities for brands and advertisers. It has been a particular boon for smaller enterprises for whom AI can produce in minutes content that previously would have been time and cost prohibitive.

But when AI produces creative assets of demonstrable value to a business, the question naturally turns to if and how these assets can be protected, and what can be done against third parties who use them without authorisation.

 

When an AI model produces an output, can this output be protected by copyright?

On this question, China has been an early global leader. Court decisions on the subsistence of copyright in AI-generated and AI-assisted content were issued as early as 2020: in one, the court held that an AI-assisted article was entitled to copyright protection and that the defendant’s publication of the article infringed the plaintiff’s rights;¹ in another, the court (and appeal court) held that the plaintiff did not enjoy copyright in certain AI-generated graphics.² However, a particularly influential decision came in 2023, in which the Beijing Internet Court held that the plaintiff’s AI-generated artwork qualified for copyright protection, the plaintiff was the copyright owner, and the defendant, in using the artwork without the plaintiff’s watermark, had infringed the plaintiff’s rights.³ This decision has proven to be influential and has been designated as a ‘typical case’ for other Chinese courts to refer to.⁴ Since it was handed down, courts in numerous other suits have agreed that copyright can subsist in AI-generated works so long as certain conditions are met (see more on this under the next heading).⁵In a sign that China’s Supreme People’s Court believes the legal position is now relatively settled, in October 2025 it stripped China’s Internet Courts (including the Beijing Internet Court) of jurisdiction over AI copyright cases, the effect of which will be to push local courts to handle such cases.

In the UK, AI-generated works may qualify under the specific provisions for copyright protection of “computer-generated works”. However, this remains untested, and a few questions arise (see under the next headings). The UK government has considered scrapping these specific provisions, but has argued that even if this were done, AI-generated works could be protected under the general rules applicable to subsistence of copyright in the UK.

As for the EU, no specific legislative provisions apply. While there is relatively little case law on the issue, a couple of cases suggest that AI-generated works can qualify for copyright protection, provided they meet the criteria for protection applicable to all copyright works, discussed under the next heading.

 

What is necessary for AI-generated works to qualify for copyright protection? Does the prompt and other user conduct matter?

Per the fundamental principles of copyright, works must meet certain conditions in order to qualify for copyright protection. The emerging jurisprudence in China, the UK and the EU has confirmed that AI-generated works are no different.

The cases from China provide that the work must be an “original” work resulting from the author’s “intellectual contribution” (more on the author under the next heading). In contrast, an output which is primarily auto-generated by AI without any “original intellectual contribution” from the author will not qualify. In the case of a user prompting a generative AI model, the court will consider how the user crafted the prompt, including the complexity and number of prompts, how the user adjusted the model parameters, as well as how the user modified the AI model’s initial outputs. As we saw from the “Spring Breeze” case, these factors were important in the court’s decision that the work was indeed an original intellectual creation. In contrast, a court in another case dismissed the plaintiff’s claim for copyright protection of certain AI-generated images. In that case, the fact that the plaintiff was unable to regenerate the images using the same prompts highlighted the inherent randomness and unpredictability of using AI models, and underscored the lack of true, author-driven expressions in the images.

In the UK, the legislation provides for the protection of “computer-generated works” provided that they still meet the requirements of originality. This is a relatively low threshold, but again, its applicability to AI-generated works has not yet been tested in court.

In the EU, it is arguable that an AI-generated work can qualify for protection so long as it complies with the general criteria of originality, the work is the author’s own intellectual creation, is an expression of their free and creative choices, and is expressed in a manner which makes it identifiable with sufficient precision and objectivity, even if such expression is not necessarily in permanent form. Similar to the position in China, the user’s sophistication in crafting a prompt and engaging with the AI model could support their claim for protection. In one 2023 decision from the Czech Republic, the court found that, in theory, a work generated by prompting an AI model could be protected by copyright, if it is the unique result of the creative activity of a natural person, and where the author can demonstrate their unique creative contribution. As this was not established on the facts, the court denied the plaintiff’s claims for protection for the work in that case.⁶ Similarly, the Italian Supreme Court in 2023 commented that if software is considered a mere tool through which a human subject creates works, the content could still be considered an expression of the intellectual creativity and personality of its author. It follows that similar remarks could apply also to AI-powered creations.⁷ A similar approach was taken in a 2026 decision of a first-instance court in Germany, which concerned AI-generated logos. The court found that copyright protection may be conceivable as a result of human intervention in AI‑generated outputs. However, they emphasized that copyright protection is not established if, in the course of prompting, the creative decision-making is left to the AI through numerous merely general and open-ended instructions.⁸

A corollary of these originality requirements is that companies which increasingly rely on AI to generate content (including, notably, software code) may find that the scope of copyright protection available to them contracts in proportion to the degree of AI involvement. Where AI tools are used to produce code or other works with minimal human creative input, it may be more difficult to demonstrate the original intellectual contribution necessary for copyright to subsist. This risk of “copyright contraction” is not confined to the UK, the EU and China, and may come into play in other territories, most notably the US.

 

If an AI-generated work qualifies for copyright protection, who is the copyright owner?

Like most other territories in the world, the UK, EU and China have all rejected the idea that an AI model can be the owner of works it generates. However, there remains a question as to who the owner should be, depending on the facts.

The courts in China have declared the user who prompted the model to be the owner, but in at least one case have declared the owner to be the model developer, demonstrating that the court will consider which entity made the relevant original intellectual contribution.⁹

The position in the EU remains untested, however a case from the Czech Republic does suggest that an AI model user could be declared the owner, provided that it can be proven that the conditions of originality indicated above are met: in particular, that the work reflects the personality of the author, as an expression of his or her creative choices, which can be shown, for instance, in the arrangements made for the creation of the output. In such cases, evidence of the prompt used might become fundamental.¹⁰

As for the UK, the legislation provides that “in the case of a literary, dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken”. In theory, this could mean that the author could be a user who prompts the model to create the work. However, this remains untested. In a 2006 case, the court held that where users had played a computer game in such a way as to create images on the screen, the person who made the necessary arrangements to create the work was the designer of the game and not the users. In the context of generative AI, this precedent gives grounds to argue that it should be the AI model developer and not the user who would be the author.¹¹

In any case, and subject to local law, the authorship and ownership position may depend on what contractual agreements are in place. For example, the user terms and conditions of many generative AI models for mass consumer use provide that copyright and other IP in content generated by the model belongs to the user.

 

Practical measures regarding copyright in AI-generated works

Although the law in this area remains in flux, several practical recommendations can be made.

• It remains possible that in many jurisdictions AI-generated works will be subject to copyright protection, giving the copyright owner rights to restrict use of such works and seek remedies for infringement. As such, AI-generated works should not be treated as “up for grabs”. Clearance should be conducted before use of such works, and in some cases, seeking licences or permissions may be necessary.
• Claimants seeking copyright protection for their own AI-generated works in multiple territories should be conscious that the law may vary significantly. This and the other articles in our series have focused on the UK, EU and China where copyright protection for AI-generated works is at least theoretically possible. However, we have not touched on the US, a territory of major global importance when it comes to AI, and one which takes a relatively strict approach to copyright protection for AI-generated works.
• When seeking to claim copyright protection in AI-generated works, evidence of the creation process will be essential in order to demonstrate the claimant’s original intellectual contribution towards the creation of the work. This evidence includes a record of what prompts were given, and what selections, modifications, choices and arrangements were made. Where work is the final outcome of an iterative process, evidence of all stages of this process should be documented. Screen recordings may be an effective way to do this.
• Businesses using AI tools to generate content at scale (including software code) should be aware that heavy reliance on AI-driven generation may weaken the copyright protection available in the resulting output. Consider whether the human contribution to the creative process is sufficient to meet the applicable originality threshold and, where it is not, what alternative protections (such as confidentiality obligations, contractual restrictions, or trade secret protection) may be available to safeguard the work.
• Some territories, including China, have a system of copyright registration in which the existence and ownership of copyright in a work can be officially documented. Under such systems, registration is generally not a formal requirement for copyright protection to arise. However, it greatly facilitates (and in some territories is necessary for) enforcement and provides evidence of the existence of copyright that is useful in the event of a dispute.
• Though not strictly related to copyright, it is worth monitoring emerging rules regarding labelling of AI-generated works, such as those in China and Italy, as well as the EU’s December 2025 draft code on transparency of AI-generated content. These rules will impact persons wishing to use AI-generated content for commercial purposes, including on traditional or social media, and in advertising or marketing.

This article was the third in our series of three articles on the status of AI and copyright issues in Q2 2026 in the UK, EU and China. The first explored whether AI training can result in copyright infringement. The second explored whether AI-generated content can result in copyright infringement. While many of the questions explored remain uncertain and indeed contentious, what is clear is that AI is a technology of generational importance, the transformative potential of which has yet to be fully seen. In the meantime, DLA Piper is monitoring the legal and regulatory landscape on an ongoing basis and is well placed to assist clients in all sectors with AI-related projects. If you have questions about anything in this article or indeed regarding AI more generally, please get in touch with the authors or your regular DLA Piper contact.