Trademark lawyer Peter Vaughan reports that the UKIPO is revising its AI guidelines following the High Court’s decision on the patent eligibility of AI machines.
Legal developments or legal changes
Dr. Thaler and his DABUS machine have sparked significant debate in patent law, particularly focusing on whether AI-generated products can attribute the AI as an inventor and, crucially, whether these outputs are eligible for patents. Surprisingly, there has been relatively limited discourse regarding the patentability of the AI machine itself. Within the UK, the prevailing approach has been to depend on the standard criteria for patent eligibility.
The previously hands-off approach is now under scrutiny. The UK High Court’s ruling in the case of Emotional Perception AI Ltd v Comptroller General of Patents, Designs, and Trade Marks [2023] EWHC 2948 (Ch) has significantly impacted the discussion on AI patent eligibility, creating a seismic shift. By eliminating the application of the computer program exemption, Sir Antony Mann’s decision has transformed the UK into a more receptive environment for patents involving artificial neural networks (ANNs) compared to just a few months ago.
Spanning 84 concise paragraphs, this case holds relevance for all ANN owners navigating patent considerations. The exclusion of the “program for a computer” exception from the patentability criteria prompted the UKIPO to swiftly issue guidance confirming that ANNs wouldn’t face objections under this exception, marking a pivotal change within weeks of the court’s decision.
In the specific case, Emotional Perception had sought to patent an ANN designed to suggest a media file—let’s assume it was a music file—to users. Post-training, this ANN generated recommendations by assessing the “closeness” or similarity between one music file and another. The concept was that the closer the files, the higher the likelihood of user enjoyment as a recommendation.
Initially, the UKIPO examiner declined the patent application, citing it as a program for a computer that lacked a technical contribution. This decision aligned with the established and rather stringent criteria applied to assess software patents.
During the appeal, Sir Antony Mann deliberated on two key inquiries. Initially, he examined whether an ANN could be classified as a program for a computer. Subsequently, if it fell under that classification, he assessed whether there existed a technical contribution.
In both aspects, Sir Antony ruled in favor of Emotional Perception. Particularly, it’s the decision regarding the former that significantly alters the UK’s perspective on the patentability of ANNs.
Firstly, he determined that neither a “hardware” ANN (essentially a closed system containing the ANN) nor a “software emulation” mimicking a hardware ANN could be categorized as a program for a computer. The ANN stood outside the boundaries of this exception. Despite human input in setting the ANN’s parameters, it operated autonomously without further intervention. Hence, an ANN did not fit within this exception, leading to the success of the appeal.
This ruling holds immense significance. It implies that practically no ANN would fall under this exception. Consequently, AI machines find themselves in a more advantageous position regarding patent eligibility compared to traditional software implementations that achieve similar outcomes. One might question whether the exception’s drafter envisioned this type of loophole, but the court isn’t the appropriate platform for the necessary policy debate.
Regarding the second aspect, Sir Antony determined that the recommendation of a (music) file to a user constituted a technical contribution. This recommendation existed outside the computer system, thereby placing the ANN beyond the exemption criteria.
This aspect of the ruling potentially extends upon previous decisions, albeit in an expansive fashion, that explore analogous outputs. It might have broader implications for computer programs aiming for patent safeguarding by suggesting that providing a file to a user could constitute a technical contribution. If the decision had primarily relied on this aspect, the seismic impact might have been more subdued. To some extent, it remains a matter specific to the case’s circumstances.
Conversely, by asserting that an ANN, whether hardware-based or emulated, doesn’t qualify as a program for a computer, a fundamental principle is set, eliminating a substantial obstacle.
The ruling has remarkably expanded the scope for ANN patents without delving into the policy considerations. However, Sir Antony’s discretion was somewhat restricted by the UKIPO’s acknowledgment that hardware-based ANNs didn’t fall under the exemption. This concession constrained the court’s flexibility, a point Sir Antony regretfully acknowledged when he remarked, “I think that the debate would have been interesting had the concession not been made.”
Additionally, the IPO was unable to argue that an ANN, at a particular level of abstraction, constitutes a mathematical model because this argument hadn’t been properly presented. This issue remains unresolved, potentially awaiting further examination.
It’s regrettable that the IPO conceded and swiftly adjusted its approach following the verdict. This case signifies the initiation of a fresh phase in the discourse concerning AI and patents. Undoubtedly, the UK has now become a more favorable environment for obtaining patent coverage for ANNs. However, a comprehensive debate from a policy standpoint, unrestricted by court confines, would be beneficial. Presently, owners of ANNs should approach UK patents with heightened interest and less apprehension than previously.
Peter Vaughan holds a position as a chartered trademark attorney and serves as a senior lecturer at Nottingham Law School.
