Protecting code and creativity: software IP rights and AI-generated code
Insight
The rise of vibe coding – the practice of using natural language prompts to create AI generated software code – is making software development more accessible than ever. Previously the preserve of people who understood programming languages (Java, Python, etc), it is now something we can all dabble in, albeit with mixed success.
At the same time, the growing use and sophistication of AI tools are challenging previously settled concepts of intellectual property (IP) rights and placing real pressure on traditional ideas of copyright and patent protection. Many businesses are now creating, using and commercialising software in ways that stretch familiar legal concepts.
This article looks at the main ways software can be protected in the UK, with particular focus on AI-developed and AI‑assisted works.
Copyright
In the UK, copyright is the main form of legal protection for software. The Copyright, Designs and Patents Act 1988 (CDPA) explicitly states that computer programs are treated as a type of 'literary work'. This means that copyright can protect not just the source and object code, but also preparatory design materials used to develop the program.
It is important to understand what copyright does, and does not, protect. Copyright protects the expression of an author’s intellectual creation, not the underlying ideas, functionality or outcomes. UK courts, drawing on retained EU case law, have clarified that protection may extend to the structure, sequence and organisation of a program. However, it does not cover the program’s functionality, its programming language or its data formats as such.
In practical terms, this means a competitor can build software that does the same thing, as long as they do not copy the protected expression of the original code. The CDPA also permits limited studying and testing of software for the purpose of achieving interoperability with other programs. While software providers frequently seek to restrict reverse engineering through contractual terms, those terms do not always fully displace statutory interoperability rights.
Unlike some other IP rights, copyright does not need to be registered in the UK. It arises automatically as soon as an original work is created. That said, good record‑keeping is essential. Businesses should keep clear records showing when software was created, who created it and on what terms. This will make it much easier to prove ownership and enforce rights if a dispute arises.
Ownership is a particularly important, and often misunderstood, issue. As a general rule, copyright created by an employee in the course of their job belongs to the employer. However, this does not apply where the work is created outside the terms of their employment (for example, as a personal project at the weekends or evenings).
Similarly, software written by external consultants or freelance developers will not belong to the commissioning business. Unless there is a written assignment of intellectual property rights from the original author to the business that commissioned the work, the author will typically own the copyright, even if they were paid to develop the code. For that reason, consultancy agreements should always include clear and robust IP assignment provisions.
The same care is needed when acquiring software: proper due diligence, backed up by appropriate warranties, is essential to ensure there is a clear chain of title.
Computer-generated works and AI
The UK is unusual in having specific legal provisions dealing with works generated by computers. In these cases, the law treats the 'author' as the person who made the arrangements necessary for the creation of the work. Depending on the circumstances and the contractual arrangements in place, that could be the programmer, the user or potentially the developer of the system itself.
There has been considerable criticism of this provision – not least the tension with the general requirement for a work to be an author's own intellectual creation to attract copyright protection. In response, and aligning itself more closely with EU and US approaches, the UK government has recently confirmed its intention to remove the statutory protection for computer‑generated works. To find out more, please see our article on UK copyright and AI.
The timing of any legislative change is currently unclear and is unlikely to be retrospective, meaning this quirk of UK law will remain relevant for some time. Even if it is removed, difficult questions will remain around whether and when AI‑generated output attracts copyright protection.
As noted above, a work will only qualify for copyright protection if it is the author's own intellectual creation. This has been interpreted as requiring a work to reflect an author's free and creative choices; works that are dictated solely by technical function, without room for creative freedom will not qualify for protection.
When humans write source code, the expressive choices made in structure, sequencing and implementation will typically involve sufficient free and creative choice to meet this threshold. Where generative AI tools are used to produce code, the analysis becomes more complex.
The key question is likely to be the extent to which a human has exercised creative control over the output – for example, through detailed prompting, iterative refinement and active selection or rejection of outputs – as opposed to merely triggering an autonomous generation process. Where a person’s involvement is too remote from the final output, the resulting code may fall outside copyright protection altogether.
The dividing line between AI-aided work (which is more likely to attract copyright protection) and purely AI-generated work (which may not) is not clear cut and will, no doubt, be tested in English courts in due course. European courts are already grappling with the issue. Earlier this year, the District Court in Munich(1) ruled that no copyright subsisted in three logos created using ChatGPT, finding that the AI tool, rather than the user, retained final creative control over the image generation process. The court characterised the human input as a set of general instructions rather than a protected act of authorship.
If the courts were to adapt a similar approach in the UK – which seems plausible given the alignment on the requirement of originality for copyright subsistence – a lot of 'vibe-coded' software would lack copyright protection. This may not always be problematic: the speed and cost benefits of AI‑assisted development may outweigh the risk in many contexts. Nonetheless, it is a risk software developers and commissioning businesses should understand and actively manage.
Patents
Patents work very differently from copyright. While it is possible to obtain patent protection for some software‑based innovations, this remains the exception rather than the rule.
Software is not automatically eligible for patent protection in the UK. The Patents Act 1977 excludes 'a program for a computer' from patentability, where the application relates to that subject matter 'as such'. The precise scope of the exclusion has been tested extensively in case law, with an important development earlier this year.
In Emotional Perception AI v Comptroller General of Patents, Designs and Trade [2026], the Supreme Court considered whether an artificial neural network constituted a patentable invention or was merely a computer program 'as such'. The decision has lowered the bar in certain contexts, but software developers still face a challenging path.
Applicants must still show that the invention is novel, involves an inventive step and makes a relevant technical contribution. Combined with the cost of securing patent protection – often exceeding £10,000 – it is easy to see why patents are rarely the first choice for protecting software.
That said, patents can be strategically valuable in an AI‑driven environment. Unlike copyright, infringement can be established without any need to demonstrate copying, or even human involvement in the development process. For businesses with genuinely innovative software solutions, patents can therefore offer a more robust enforcement tool.
For businesses thinking about patenting a software innovation, it is sensible to carry out patent clearance searches early in the development process. This helps to confirm whether there is freedom to operate before significant investment is made, and it may also highlight opportunities for defensive patent filings.
Database rights
Where software involves collecting, organising or presenting large amounts of data, database rights under the Copyright and Rights in Databases Regulations 1997 may also be relevant. These rights protect against the extraction or reuse of substantial parts of a database where there has been substantial investment in obtaining, verifying or presenting the data. Database rights can be particularly valuable for products such as analytics platforms, data management tools and AI training datasets.
IP infringement risks
In addition to the above questions over the existence and ownership of copyright, using generative AI tools such as GitHub Copilot or Cursor in software development brings specific IP infringement risks that may need to be managed. These tools are trained on large datasets that may include third‑party copyrighted code. As a result, there is a risk that AI‑generated outputs could reproduce code protected by copyright or subject to open‑source licences.
To manage this, businesses should have clear internal policies on how AI tools can be used in development. Those policies should require AI‑generated code to be reviewed and checked for potential third‑party IP issues before it is incorporated into commercial products.
Regular IP audits are also an effective way to spot problems early. A well‑designed audit should look at development processes, the tools being used, the licensing terms for third‑party software and AI platforms, and the business’s own chain of title to its software. Taking these steps early can help prevent costly disputes later on.
If you have any questions, please contact your usual contact in the IP & Commercial Team.
Many thanks to trainee solicitor Perihan Tur for her help in writing this article.
This publication is a general summary of the law. It should not replace legal advice tailored to your specific circumstances.
© Farrer & Co LLP, May 2026
