Can You Patent an Algorithm in the UK?
Generally, you can’t patent an algorithm, but it might be a practical application of an algorithm. English courts have consistently held that different approaches often lead to the same result in practice, as Birss J recently confirmed in Lenovo (Singapore) PTE Ltd v Comptroller General of Patents. But the UK courts are more flexible, so the question of whether an algorithm can be patented in the UK remains a subject for a debate between AI and patent lawyers.
Computer programs and mathematical methods are excluded from patents in the uk
UK law follows the European Patent Convention (EPC) which states that a patent should not be granted for an invention containing computer programs or mathematical methods. This excludes inventions containing mental acts and business methods. This approach has long challenged legal minds, but the UK courts have finally agreed that the presence of such categories is not determinative. The four-part test has been applied in the UK Patents Court since 2005.
The UK Patent Office (UKIPO) has said that a computer program is not an ‘invention’ if it does not have a novel effect outside the computer. This exclusion was enacted to ensure that patent law does not interfere with doctors, who often develop innovative ways to treat patients and diagnose problems. This exclusion, however, does not apply to drugs. Therefore, it is very important to know exactly what is included and excluded in order to avoid confusion.
The EPC and the Patents Act have a nexus between computer programs and mathematical methods. Both the EPO and the UK patent office have said that computer programs and mathematical methods are excluded from patents in the UK. This is not a legal precedent in itself, but it provides a useful framework for future decision-making. The EPO’s decision in Fujitsu, a recent case involving a telecom company, reaffirmed the EPO’s position.
This decision reaffirms the rule that mental acts are not patentable. The EPO Technical Boards of Appeal have interpreted this rule differently than the UK Patent Office. The UK Patent Office, however, has not followed this ruling. The UK Patent Office and Court of Appeal have not followed this precedent. Therefore, computer programs and mathematical methods will remain patentable as long as they are technical in nature.
Although the EPO rejects computer programs as ineligible for patent protection, the UKIPO accepts software-implemented methods as inventions. In the UK, however, UKIPO considers any feature that contributes to the inventive step. However, the UKPO is more conservative than the EPO. Consequently, a patent granted to a software-implemented method is more likely to be invalidated in the UK than in the EPO.
Another important change is in the EPC. This ruling makes it harder to patent computer programs. While software and mathematical methods are considered to be generic, they can still be patentable if they involve an inventive step. Moreover, the EPC has re-defined “invention” to include business methods and software. The new rules will clarify the defining criteria of an invention. The EPO has made a number of decisions that made software and business methods patentable.
Practical application of an algorithm may be patentable in the uk
A practical application of an algorithm may be patentable in England, despite European and American laws. The UKIPO does not require the disclosure of prior art in patent applications. In addition to reviewing search results from the UKIPO, the patent office will consult the patent’s designated countries’ IP offices, including those of the inventor. This may be the result of the UKIPO’s judicial review process.
While AI presents several challenges for the patent system, the UKIPO is considering a consultation process. The UK-IPO is seeking opinions on how the UK should approach AI patents. The patent system has roots in mechanical and steam engines. An AI algorithm may adapt based on data and produce a new, innovative feature in a product. The UK-IPO hopes to publish a position paper in 2022 and is now inviting views on the topic.
Problems with enforcing AI patents
The current practice of enforcing AI patents in Europe seems fairly consistent, but the UK is one of the few EU member states that does not have patent exclusions. The UK government is working to fix this problem by publishing enhanced guidelines in the area of patent exclusions in the near future. In the meantime, it is important to note that AI patents are not yet legally protected under EU law, so enforcing them in the UK could present challenges.
Another problem with enforcing AI patents is that it is difficult to prove an infringement of an AI patent. This is due to the fact that the software in question uses data to improve efficiency. This ability to learn by iteration raises interesting issues in the area of patent infringement. These issues are particularly important if the AI software is developed for use in a commercial setting.
The UK Patents Act 1977 is quite clear that a “person” must be the actual deviser of an invention in order to be liable for infringement. However, there are many problems associated with proving infringement of an AI patent – a problem that already exists with other technologies. However, AI patents are still deemed to have adequate flexibility, and claimants can use the Court process to support their actions.
Another problem with enforcing AI patents relates to the fact that many companies use the AI technology. One company may develop the AI software, which another company then trains. A third company may then purchase the trained model and use it to perform a task. The system may change incrementally, and infringement rules may not be applicable to the software used by these companies. Therefore, the UK courts may be hesitant to enforce AI patents.
Another problem with enforcing AI patents is that many countries do not have adequate protection for machine-generated training datasets. Therefore, the UK government may soon introduce a system of deposit for AI data. This would make it easier to train AI systems and would not be overly burdensome. However, this system may not be entirely practical, and may not be sufficient to address the problem.
In the UK, patent offices must consider whether an AI system is an “inventor” under current patent law. This may pose a problem in some jurisdictions, as AI patents have no human inventor. The United States Patent and Trademark Office ruled that AI systems are not ‘persons’ and cannot be the inventors of a patent. While the UK Patent Office has adopted a requirement that human inventors name themselves as inventors, this does not allow an AI system to be named as the ‘person’ of an invention.
While the patent system can be a positive factor for AI research and development, it is important to be cautious. Patents should be fair and balanced. Different modes of AI are better suited to a particular patent system than others. However, patents must be adapted to fit with the IP landscape. This is why patents must be flexible enough to accommodate the development of AI technologies. However, a small group of private individuals are arguing that the current patent system may handicap future innovation.