Having addressed copyright infringement risks, we turn to patent protection.
Patent law raises two key questions:
- can an AI system be named as an inventor; and
- can AI-assisted inventions be patented?
In short: no, and increasingly yes, subject to important limitations.
AI cannot be an inventor
In Thaler v Comptroller-General of Patents [2023] UKSC 49, the UK Supreme Court held that an AI system cannot be an inventor under the Patents Act 1977, and that ownership of an AI machine does not entitle the owner to apply for a patent over what it produces. The European Patent Office and the US Patent and Trademark Office have reached the same conclusion. South Africa's Patents Act 57 of 1978, which defines an inventor by reference to a "person", is consistent with this global consensus.
The practical consequence is significant. If your firm develops a novel AI-driven financial process, the human or humans who made the creative and technical contribution resulting in the invention must be identified and named, and their role in the inventive process must be clear. Where that documentation does not exist, patent rights may be lost.
A significant shift on patentability
Until recently, the UK took a different approach to the patentability of AI-related inventions than the European Patent Office. The UK Supreme Court resolved that divergence in Emotional Perception AI Ltd v Comptroller-General of Patents [2026] UKSC 3, holding that an artificial neural network does not fall within the excluded category of computer programs "as such", and can therefore, in principle, be patented.
The UK has effectively aligned with the European Patent Office's more permissive approach. Firms that previously received advice that AI-related inventions were unpatentable should revisit that advice.
Patentability is only the first step
Clearing the patentability threshold does not guarantee protection. Novelty and inventive steps are still required. The European Patent Office's Reinforcement Learning/BOSCH decision illustrates the point: AI-specific features that do not produce a genuine technical effect will not contribute to inventive steps, and an application will fail on obviousness grounds, even where the subject matter is, in principle, patentable.
The prior art in AI is expanding rapidly. The window for protecting genuinely novel AI-driven inventions is narrowing.
What to do now
A patent audit should be conducted across AI‑driven processes and tools within the organisation. Particular attention should be paid to areas such as credit scoring, fraud detection, algorithmic trading and regulatory reporting, where AI is often embedded in core commercial and compliance functions and may give rise to protectable inventions.
Human inventorship must be documented meticulously. For every potentially patentable invention developed with the assistance of AI, clear and contemporaneous records should be kept detailing the human creative input, technical decision‑making and problem‑solving contribution. This documentation is increasingly critical in establishing valid inventorship.
Timely action is essential. Once an invention is publicly disclosed, whether through a product launch, a regulatory filing or a conference presentation, the opportunity to secure patent protection may be lost or significantly limited. Early identification and filing are therefore key.
Finally, organisations should seek fresh legal advice considering the Emotional Perception decision. The legal landscape around AI and patentability has shifted, and strategies that were effective in the past may no longer be sufficient.
