Does copyright subsist in prompts used with large language models?
Large language models now sit at the centre of creative and commercial activity. Text is drafted, refined and repurposed at speed using generative systems that respond to human prompts. This has given rise to a deceptively simple legal question. Does copyright exist in AI prompts under UK law?
Answering that question requires stepping back from the technology and returning to core copyright principles.
Copyright fundamentals and the role of human authorship
UK copyright law protects original literary, dramatic, musical and artistic works. Originality requires the author’s own intellectual creation. That test focuses on human judgement, skill and creative choice.Human creativity is at the core of UK copyright law.
Copyright does not arise simply because a person is involved in a process. It protects expression, not activity. Typing, clicking or submitting information does not automatically amount to authorship in copyright terms.
This distinction is critical in the context of large language models (LLMs) . A human is always physically present when a prompt is entered. However, physical presence is not the legal test. The question is whether the human author determined the expressive elements of the work.
Prompts as potential copyright works
Under UK law, a prompt can itself attract copyright protection if it is sufficiently original. A short or purely functional instruction is unlikely to qualify. By contrast, a carefully structured and expressive prompt that reflects creative choices in language, tone or narrative may constitute a literary work.
This is not controversial. UK copyright law already protects written instructions, manuals and technical material where originality is present. The fact that a prompt is written for use with an AI system does not disqualify it from protection.
However, copyright in a prompt does not automatically extend to the output generated in response to that prompt. This distinction is often misunderstood.
Outputs and why prompting alone may fall short
Large language models generate outputs through probabilistic processes that are not transparent to users. The same prompt can produce different results. The same user can receive different outputs on different occasions.
That unpredictability weakens the causal link between the human author and the final expression. In many cases, a prompt operates at a high level of abstraction. It sets parameters rather than determining the final form of the text.
For this reason, prompting alone will often be insufficient to establish copyright in the resulting output. The difficulty is not that the human is absent. It is that the expressive outcome may be determined by the system’s internal processes rather than by the human’s creative decisions.
This is why lawyers and regulators resist the idea that every AI user automatically becomes the author of AI generated content.
Computer generated works and the UK exception
UK law contains a distinctive provision dealing with computer generated works. Section 9(3) of the Copyright, Designs and Patents Act 1988 provides that where a work is generated by a computer in circumstances such that there is no human author, the author is taken to be the person who made the arrangements necessary for its creation.
These works attract a shorter copyright term of 50 years from the end of the year in which the work is made. This operates as an exception to the general rule of 70 years after the author’s death.
This framework does not mean that anyone who inputs a prompt automatically becomes the author. The phrase “arrangements necessary” is narrower. It points to responsibility for the creative process as a whole rather than mere interaction with a system.
In the context of AI, that responsibility may include configuring the system, selecting and assembling outputs or shaping material into a final form. It does not clearly extend to casual or generic prompting.
Human creative control remains decisive
Where a human uses an AI system as a drafting tool rather than an autonomous creator, the legal position improves significantly.
If a user develops prompts iteratively, selects between outputs, edits text, restructures content or combines multiple outputs into a final work, those human authored elements are far more likely to attract copyright protection.
In such cases, copyright may subsist in the work as a whole or at least in the human contributions embedded within it. The more the AI behaves like a tool, the stronger the argument for human authorship.
Practical guidance: documenting human authorship in AI assisted work – the crux of copyright in Prompts
The UK has not yet tested these issues in court. There is no binding authority on whether prompts alone are sufficient to ground copyright in AI generated outputs.
As a result, legal advisers tend to recommend caution rather than reliance on automatic protection. The focus is on evidence.
In practice, creators are advised to keep records showing:
- how prompts were conceived and developed and by whom
- why particular wording or stylistic choices were selected
- how outputs were reviewed and refined
- which outputs were rejected and why
- what edits rewrites or structural changes were made by a human
- how multiple outputs were combined into a final work
This material helps establish that the AI system functioned as a tool rather than an autonomous author. It also assists in identifying the person who made the arrangements necessary for the creation of the work under the Copyright, Designs and Patents Act 1988.
Until courts provide clearer guidance, careful documentation remains one of the most effective ways to manage legal risk in this area.
Prompt & LLM Consultancies
The growing market for prompt consultancy reflects commercial value rather than legal certainty. While some prompts may qualify for copyright protection, much of what is traded consists of know how, workflow design and optimisation techniques that sit outside copyright and are better protected, if at all, through contract and confidentiality.
Policy context and future direction
The UK Intellectual Property Office has acknowledged that artificial intelligence challenges core copyright assumptions. Consultations have explored whether existing protections for computer generated works remain appropriate and whether reform is required.
For now, the UK position remains cautious and incremental. AI systems are treated as tools rather than authors. Human creativity remains central. Prompts can be protected where they are original. Outputs require demonstrable human authorship to attract reliable protection.
Recent US Copyright Office guidance on prompts/ Breaking: Update from US Copyright office
Recent guidance from the US Copyright Office on January 29th of this year, just confirmed that prompts alone will not normally support copyright in AI generated outputs under US law. While this guidance has no direct legal force in the UK, it offers another approach to AI-related copyright legislation. Both UK and US law emphasise that copyright protection turns on human authorship expressed in the final work, not on interaction with a system as such.
The UK approach differs structurally, particularly because of section 9(3) of the Copyright, Designs and Patents Act 1988. However, the underlying emphasis on human intellectual creation is consistent. UK courts are therefore unlikely to treat prompting alone as sufficient where the expressive outcome cannot be traced to human creative control.
Conclusion
Copyright can exist in AI prompts under UK law, but only where those prompts reflect original intellectual creation. Copyright in AI generated outputs is far less certain and will usually depend on the extent of human creative control exercised beyond prompting alone.
The presence of a human at the keyboard is not enough. Authorship still turns on intellectual creation.
Taylor Hampton’s Copyright Practice
As a leading boutique media law firm, we act for individuals, businesses and celebrities in the entertainment arena. Our experience working with the UK Press brings many years of knowledge in the law in this sector. For more information on Taylor Hampton’s copyright practice see HERE or contact 00442074275970.
