A Legal Reset for AI Patents; The Turning Point for the AI and Software Patents Race in the UK

The boundaries of patent law are once again being tested by artificial intelligence. In a recent decision, the UK Intellectual Property Office (IPO) refused a patent application from AI music company DAACI for a system designed to automatically generate music based on an emotional brief. Despite the system’s ability to produce compositions with “audibly correct form,” the office concluded that the invention amounted to nothing more than a “program for a computer”, a category long excluded from patent protection under UK law.

The ruling highlights a growing legal tension: as generative AI systems increasingly blur the line between technical innovation and creative output, patent law is being forced to decide where software ends and invention begins.

At the centre of the dispute was a patent application filed by DAACI in March 2021 for a system designed to automatically generate musical compositions. The invention, described in patent GB 2605440A, aimed to create music that responds to an “emotional brief” provided by a user. In practice, the system would interpret narrative or emotional cues, such as tension, excitement, or melancholy and generate a composition intended to reflect those moods while maintaining what the company described as an “audibly correct form.”

The very nature of the invention; an AI system designed to automate the creative process, would become central to the legal challenge that followed. When the application came before the UK IPO, the key question was not whether the system could generate convincing music, but whether the underlying innovation amounted to a technical invention at all.

In assessing DAACI’s application, the UK IPO relied on the long-standing Aerotel framework for determining whether an invention falls within excluded subject matter such as a “computer program as such.” Under that approach, decision-makers first identify the invention’s contribution and then assess whether that contribution is technical in nature. Applying this reasoning, the hearing officer accepted that DAACI’s system contributed something new by automating musical composition in response to emotional prompts. However, he concluded that the problem being solved, producing aesthetically meaningful music, was ultimately creative rather than technical, meaning the invention remained within the exclusion.

However, just days after the DAACI decision was handed down, the Supreme Court ruled in Emotional Perception AI Ltd v Comptroller General of Patents that the Aerotel test should no longer be followed. Instead, UK law must align with the approach of the European Patent Office (EPO), particularly the Enlarged Board of Appeal’s decision in G1/19.

Under the new approach followed by the EPO, the first hurdle is relatively easy to clear. If an invention involves some form of technical system, such as software running on hardware, it will usually count as an “invention” in principle. But that does not mean it will receive a patent. The real test comes next, when examiners look closely at the invention’s features and ask a tougher question: which parts of this system actually represent a technical improvement? Only those technical elements are considered when deciding whether the invention is truly new and inventive compared to what already exists. This matters because, as AI systems become more sophisticated, whether an invention passes this test could determine which companies can claim ownership of the underlying technology, and ultimately shape who leads the next generation of AI-driven creativity.

Viewed through this updated lens, the reasoning in DAACI appears somewhat out of step with the direction of UK patent law. The new framework would focus more narrowly on whether the claimed system involved technical means and which elements of it possess genuine technical character. For AI-driven systems operating at the intersection of software and creativity, that shift could prove decisive.

The timing of the DAACI decision makes this tension particularly striking. The removal of the old Aerotel framework and the shift toward the EPO’s approach may make it easier for AI inventions to pass the first hurdle of patent eligibility. But that does not mean patents for generative AI will suddenly become commonplace. Instead, the battleground is likely to move deeper into the analysis, where companies must show that the technical machinery behind the AI, its architecture, training processes, or data processing methods, represents a genuine technological advance.

This shift could also change how AI developers approach patent strategy. Rather than emphasising the creative results produced by their systems, companies may increasingly frame their inventions around the technical processes that enable those results, for example; the way an algorithm processes input data, structures musical elements, or optimises the generation of compositions. In other words, the legal focus may move away from the music itself and toward the technical infrastructure that allows machines to create it.

If patents do begin to emerge around these kinds of AI systems, the effects could be far-reaching. AI models and training techniques could become some of the most valuable intellectual property in the technology sector, attracting investment but also raising the stakes in an already intense race to develop more powerful systems. At the same time, stronger patent protection could reshape competition, potentially allowing early movers to lock up key technological building blocks.

Ultimately, cases like DAACI reveal how uneasy the relationship between creativity and patent law has become. AI may be capable of composing music at the touch of a button, but the law still insists on asking the harder question: is the invention behind it is truly technical? And as AI grows more capable, that question is only going to get harder.

© Lawrence Power 2026

Bibliography:

Leave a Reply

Your email address will not be published. Required fields are marked *