As courts and copyright offices around the world grapple with whether—and how—to protect AI-generated and AI-assisted works, the United Kingdom finds itself in an unusual position. Long before the advent of large language models and image-generating algorithms, UK law explicitly acknowledged the possibility of a copyrightable work being created without a human author.
This blog post reflects on the legislative history and jurisprudence behind Section 9(3) of the Copyright, Designs and Patents Act 1988 (CDPA)—a provision that purports to offer copyright protection for computer-generated works. In my view, whether Section 9(3) is suitable for protecting outputs generated by contemporary generative AI models—capable of producing sophisticated outputs with minimal human involvement—is far from clear and is now under active review in the UK.
UK copyright law has traditionally required a human author to satisfy the originality requirement. This changed, at least in theory, with the introduction of Section 9(3), which allows authorship to be assigned by legal fiction where no human creator exists—enabling such works to meet the basic conditions for protection. Section 9(3) of the CDPA addresses a particular class of works referred to as “computer-generated works” (CGWs). The statute provides:
“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.”
The term “computer-generated” is defined in Section 178 of the CDPA as referring to works “generated by a computer in circumstances such that there is no human author of the work.”
If these provisions apply, copyright subsists for 50 years from the end of the year in which the work was created. The “author” is assigned by legal fiction to the person making the necessary arrangements, and such imputed “authors” do not receive any moral rights.
This framework was enacted in 1988 and originally aimed at outputs from early computer systems. At the time of its introduction, Section 9(3) was the product of lobbying by stakeholders in the nascent software and computing industries. Industry participants were concerned that the absence of copyright protection for non-human outputs under the previous 1956 Act posed a risk to investment. If not protectable by copyright, investors feared that such works could be freely copied by competitors, undermining the commercial value of the technology and disincentivizing further innovation. As such, the provision was designed in part to reassure investors that computer-generated outputs would be protectable assets, thereby encouraging technological development and commercial certainty in the sector.
However, despite being on the statute books for over three decades, Section 9(3) CDPA has received remarkably little judicial interpretation. In fact, with only two cases of note, there is a limit to how much doctrinal clarity one can extract from the case law. The provision has only been ruled on by a court once —in Nova Productions Ltd v Mazooma Games Ltd, where the High Court held that images generated during gameplay were protectable computer-generated works, and that the programmer was the author for the purposes of Section 9(3), because he had written the code and designed the rules governing how the images were produced. The player, though instrumental in causing the images to appear, was not considered to have made the “arrangements necessary” for the work’s creation.
In contrast, in THJ Systems Ltd v Sheridan, both parties had advanced claims of authorship under Section 9(3) in the pleadings. The case concerned a dispute over who held copyright in financial charts generated by software—specifically, whether the rights belonged to the developer of the computer program or the individual who had used it. However, the Court of Appeal did not interpret Section 9(3) at all. Instead, it focused exclusively on the standard of originality defined post-Infopaq as involving the exercise of “free and creative choices” and the presence of a “personal touch”, concluding that the software developer had made sufficient free and creative choices—such as in the selection of colours and font styles in the visual charts—to be deemed the human author under conventional copyright principles. Once this was established, there was no analytical need to consider Section 9(3).
These decisions appear to reflect an emerging judicial tendency to locate traditional human authorship wherever possible. Section 9(3) appears to operate more as a fallback in practice: courts begin by assessing whether a human can satisfy the originality requirement under Section 1(1)(a), and only turn to the legal fiction in Section 9(3) if no such author can be identified. This approach arguably demonstrates a reluctance to engage with the interpretive uncertainties surrounding Section 9(3), particularly given its sparse case law and doctrinal ambiguity.
Where a work is wholly generated by a computer without such human input, Section 9(3) assigns authorship by legal fiction to the person who made the necessary arrangements in the computer program. The CDPA constructs a causal link between the work and the person responsible for its creation to fill the authorial void. This attribution may best be understood as a policy mechanism designed to support investment, not as recognition of personal creative expression.
This distinction between traditional human authorship and Section 9(3) may inform our understanding of the current debate over AI-assisted works and computer-generated works. AI-assisted works are those where a human uses AI as a creative tool. They are protected under traditional authorship rules, the human user is the author, and the work must meet the originality threshold of embodying “free and creative choices” and a “personal touch.” Moral rights apply to such works. In contrast, computer-generated works involve no human author at the point of creation. They are governed by Section 9(3), the author is assigned by legal fiction, and originality is presumed to be satisfied. Moral rights do not apply to such authors.
In light of rapid technological developments, the UK government and the Intellectual Property Office (UKIPO) have begun reassessing whether Section 9(3) still serves a meaningful purpose. In its 2024 statement, the government suggested that the provision might be repealed unless a strong case for its retention is made. I believe this reflects a growing discomfort with the doctrinal ambiguity surrounding Section 9(3), particularly in light of its tension with the current standard of originality. While it is true that AI-generated outputs that have been litigated in UK courts so far have typically been found to involve at least some human input (and thus fall under conventional authorship rules) this doesn’t necessarily mean that Section 9(3) is irrelevant. Rather, it reflects how the law currently leaves little room for outputs that lack human involvement to satisfy the originality requirement. Ironically, that very limitation points to why Section 9(3) was created in the first place: to fill the gap for works that defy traditional authorship frameworks.
The government’s statement acknowledges that Section 9(3) could still serve a purpose—particularly for works generated “in response to a simple prompt,” which may not meet originality thresholds but are nonetheless commercially or culturally significant. Yet without reform, it is worrying that the provision will continue to exist in a kind of legal limbo: too uncertain to rely on, yet too conceptually important to discard. Compounding the problem is the lack of publicly available data on how Section 9(3) is used in practice. This is perhaps unsurprising given the UK’s lack of a formal copyright registration system, but the lack of transparency has complicated efforts to assess the provision’s real-world use and has fueled arguments that the section is either underutilized or insufficiently understood by rights holders.
Unlike the UK, the United States maintains a firm human authorship requirement. The U.S. Copyright Office has consistently refused to register works lacking human authorship, such as AI-generated images in Zarya of the Dawn. Nor do the U.S. Copyright Office or courts recognize machines as authors, as illustrated in Thaler v. Perlmutter. As a result, the UK’s statutory approach may seem attractive to U.S.-based creators seeking protection for fully AI-generated outputs. However, such protection would be geographically limited. U.S. courts apply domestic copyrightability standards even to foreign works, and a work protected under UK law is not guaranteed recognition in the U.S. absent human authorship.
Section 9(3) of the CDPA remains one of the few legislative attempts to grapple directly with the question of authorship in the absence of a human creator. In theory, it reflects a bold and pragmatic effort to address a problem that many jurisdictions are only now beginning to face. But in practice, it reveals the growing tension between older legal frameworks and the realities of contemporary AI creation. By assigning authorship through legal fiction, Section 9(3) was designed to preserve copyright’s economic function—namely, to incentivize investment—even when personal creativity is absent. That rationale still holds value, especially in light of generative AI’s increasing ability to produce culturally and commercially significant works. But without judicial guidance or doctrinal clarity, the provision now risks becoming more symbolic than functional. Its rare application in case law, the judiciary’s reluctance to invoke it, and the government’s own doubts about its ongoing relevance all point to a deeper problem: the law remains uncertain about how to handle machine authorship. For that reason, I would argue that Section 9(3) should not be viewed simply as a legislative relic—but also not as a ready-made solution. Instead, it should serve as an invitation to rethink how we define authorship, originality, and protection as human and machine creativity become increasingly intertwined.
Discover more from Authors Alliance
Subscribe to get the latest posts sent to your email.