AI cannot be an inventor, but is reform on the horizon?
Like many other IP offices, the UKIPO has been carefully considering the role of patents in promoting innovation in artificial intelligence (AI), and the use of AI in the innovation process. Following a call for views on how the UK's IP framework caters for the development and wider use of AI, the Government has recently published its response. As part of this process, the Government intends to consult later in 2021 on a range of possible policy options, which may see revisions to the Patents Act 1977, in relation to AI generated inventions which would otherwise not meet the test for protection. The UKIPO will also publish enhanced guidelines on patent exclusion practice for AI inventions, and commission an economic study to provide a basis to judge further intervention.
The call for reviews follows the Patents Court decision in Stephen L Thaler v The Comptroller-General of Patents, Designs and Trade Marks that AI systems cannot, as things currently stand within the scope of UK patent legislation, be acknowledged as an inventor. The Court's decision concerned the 'DABUS' AI machine, and focused on the requirement that, pursuant to the Patents Act, an inventor is a 'natural person'. As well as the UKIPO, both the EPO and the USPTO also refused patent applications based on DABUS as the inventor, with a number of other applications having been made on a worldwide basis.
Whilst the Court recognised the importance of the debate about what the law should say about AI, it emphasised that this is a complex policy issue. The Court's role was only to construe legislation; it could not itself legislate, no matter how great the policy need. The case is listed before the Court of Appeal in July 2021.
Supreme Court finds transgenic mice patents invalid for insufficiency
A patent represents a bargain between the patentee and the public - in return for a valuable 20 year monopoly right over use of the invention, the patentee discloses the invention to the world, thereby offering the ability to others to 'work the invention' after the monopoly has expired. In order to patent an inventive product, patent law requires that the skilled person in the art would be able to make the product by using the teaching disclosed in the patent, together with the common general knowledge available at the priority date, without undue experimental burden or having any inventiveness.
In Regeneron v Kymab, the Supreme Court described this sufficiency requirement as a 'bedrock' of patent law, which prevents patentees obtaining a monopoly exceeding their contribution to the art. In particular, the Supreme Court had to consider sufficiency in circumstances where the invention was for a range of products – here, a ground-breaking range of types of genetically modified mouse – but the disclosure in the patent only enabled the skilled person to make some of the types of product in the range at the priority date.
The Court of Appeal, which found the patents valid, concluded that there was no need for Regeneron to disclose how to make the full range of mice because the idea was a "principle of general application".
The Supreme Court overturned this decision, whilst recognising that the Court of Appeal's approach would increase the rewards for inventors in complex, rapidly developing areas like genetic engineering for treating diseases. The Supreme Court majority concluded that the relevant claims 'clearly' failed for insufficiency. At the priority date, the disclosure of the patents, together with the common general knowledge, did not enable transgenic mice at the more valuable end of the range to be made. Therefore the claim to a monopoly across the whole of that range went far beyond the contribution that the product made to the art at the priority date. The Court of Appeal's approach had watered down the sufficiency requirement and had tilted the careful balance between a patentee and the public in an unwarranted way.
The Patents Court has recently applied the Supreme Court's decision in Illumina v Latvia MGI & ors (discussed here). As the application of Regeneron evolves in the Patents Court it is inevitable that we will see disputes concerning “Regeneron insufficiency” revolve around arguments over what is or is not the essence or core of the invention and whether the variable significantly affects the value or utility of the claimed product or process.
Earlier obscure 'paper proposal' invalidates hose patent
As a multi-factorial evaluation, the Court of Appeal rarely intervenes with a trial judge's assessment of whether a patented invention involves an inventive step, unless it identifies an error of principle or law. This includes those cases where the result may, as the Court accepted in Mishan v Hozelock a case relating to expandable garden hoses, "appear harsh, even unfair" to the inventor. Whilst there was an acknowledgement in this case of a feeling that the inventor should be able to "reap the rewards of his brilliantly simple invention", the wider public interest required that patent protection should not be available where there is prior art (no matter how obscure) either disclosing the invention or making it obvious. As such, the patent was held to be invalid.
The decision is noteworthy for a disagreement on the approach to take on a key issue between two highly experienced patent judges: Arnold LJ, who was elevated to the Court of Appeal from the Patents Court in 2019, and Floyd LJ. Giving the majority judgment, Arnold LJ noted that the trial judge had asked himself, correctly, whether the differences between the prior art and the inventive concept were steps which would have been obvious to the skilled person "viewed without any knowledge of the alleged invention as claimed" and had expressly stated that he did not consider his conclusion was based on hindsight. The majority therefore concluded that the finding of obviousness was one that the trial judge could properly make: the prior art unambiguously disclosed the same basic concept of an extendible and retractable hose as the invention claimed in the Patents. However, Floyd LJ disagreed, highlighting the particular danger of hindsight when considering a simple invention such as this. In his view, as a "mere paper proposal", the prior art was less likely to render the invention obvious.
Akebia v Fibrogen: Broad claims found invalid on various grounds
In a lengthy judgment dealing with a number of issues, Arnold LJ (on return to the Patents Court) held that six Fibrogen patents (exclusively licensed to Astellas) concerning the use of enzyme inhibitors for treatment or prevention of anaemia and related conditions were invalid (apart from one claim). The patents were held invalid variously on the grounds of obviousness and/or insufficiency for lack of plausibility, undue burden and/or uncertainty. The number of compounds in the claims covered by a formula in the patents was described by one expert witness in the case, with whom Arnold LJ agreed, as "staggeringly large". Interestingly, in relation to a claim that was held to be valid, Arnold LJ rejected a claim to infringement on the basis of equivalence. The case is listed before the Court of Appeal in June 2021.
Merck v Wyeth: Vaccine formulation patent invalid due to obviousness
Wyeth owned a patent for a formulation of vaccine which prevented silicone-induced aggregation by use of a surfactant. Merck claimed that the patent was invalid due to obviousness on the basis of a paper published in Spanish in 2004 by Wyeth’s workers in Madrid concerning pneumococcal vaccination in Spain.
The Patents Court held that the patent was obvious over the prior art and therefore invalid. Although the prior art was of primary interest in Spain, it was not limited to Spain and so it was appropriate to consider in this case as, even though it probably had limited readership, this did not mean it would not have been read by a skilled vaccionologist (the skilled person in the art in this case). The Court held that the claims of the patent took forward an attractive proposal contained in the prior art by routine means, including solving a modest common general knowledge problem (the aggregation caused by silicone) in a way which was already contained in the common general knowledge (use of a surfactant). The case is listed before the Court of Appeal in July 2021.
Lufthansa v Astronics: Extent to which prior disclosure referenced in patent should be used in construction
This case concerned Lufthansa's patent for an apparatus which provided a supply voltage for electrical devices in aeroplane cabins. Lufthansa alleged that Astronics had infringed its patent and Astronics counter-claimed that the patent was invalid on the basis of novelty and lacking an inventive step. The Court, therefore, had to consider the prior art in this case.
The Court held that the patent was valid and infringed. When considering validity, the Court considered a prior art patent, which was specifically described in the specification of Lufthansa's patent, and whether the skilled person was expected to have read and understood that prior art patent. Morgan J held that "there is no principle of law which requires a finding that the skilled addressee of the Patent would obtain and consider the full patent specification in [the prior art patent] and then take it into account when construing the Patent". Essentially the patent could be construed on its face together with the common general knowledge of the skilled person without reference to the prior art. This was despite the German Courts having considered the full specification of the prior art in a case concerning a German equivalent of the patent in issue. However, Morgan J held that that was "not the right approach in this jurisdiction in the circumstances of the present case". The case is listed before the Court of Appeal in November 2021.
Edwards v Meril: Patent for heart valve invalid due to obviousness and patent for heart valve delivery system partially invalid and infringed
Edwards owned patents for a prosthetic heart valve and a heart valve delivery system. Meril sold a transcatheter heart valve and a catheter based delivery system for the valve which Edwards claimed infringed its patents. Edwards also relied on the doctrine of equivalents in relation to its delivery system patent. Meril denied infringement and counterclaimed for revocation of the patents on the grounds of, amongst other things, obviousness due to a prior patent and patent application.
Birss J held that the heart valve patent and some claims of the delivery system patent were invalid due to obviousness as a result of the prior art relied on by Meril which disclosed embodiments of Edwards' previous heart valve system. However, Birss J held that Meril's delivery system product did infringe the remaining valid claims of Edwards' delivery system patent both on the basis of the normal interpretation of the claims and the doctrine of equivalents.
Another notable feature of this case is that Birss J agreed to conduct the trial as a 'hybrid hearing' whereby some participants attended the Court in person and some via video link. The Court was prepared to do so in circumstances where it was carried out with the Court's permission and under the Court's control. It also promoted access to justice and was necessary for a fair trial since the physical attendance by overseas participants may otherwise have been impossible (due to COVID-19 travel restrictions).
Eli Lilly v Genentech: Validity case concerning divisional patent where issues of abuse of process and issue estoppel arose
In March 2019, Genentech's patent for antibodies was found to be invalid for obviousness and insufficiency. However, Genentech then obtained a divisional patent which concerned materially the same subject matter. Eli Lilly applied for summary judgment that the divisional patent be revoked on the basis of an abuse of process and issue estoppel because the parent patent had been found to be invalid.
An abuse of process (also known as the Henderson v Henderson principle) arises where a party seeks to raise in subsequent proceedings matters which were not raised in earlier proceedings but could, and should, have been.
An issue estoppel will only arise where:
- An issue has been determined in a final decision in an earlier action between the same parties;
- The issue determined in the earlier action must be fundamental to the earlier decision; and
- The issue in the later action is the same as was finally determined in the earlier action.
The public policy rationale behind these principles is to ensure finality for parties in litigation and that a party should not have to litigate the same matter twice.
In February 2020, the Patents Court held that Genentech was estopped from arguing that claims 1-9 of the divisional patent were valid because they were attempting to re-litigate a question of validity which had already been decided in respect of the parent patent (which contained very similar claims) in 2019. Factors the Court took into consideration when reaching this conclusion included the fact that the previous proceedings were "one of the most complicated and expensive patent actions" and that Genentech did not have its own product on the market but was seeking to keep Eli Lilly's product off the market. However, the Court held that Genentech had a real prospect of arguing that the remaining claims 10 and 11 in the divisional patent. As such, the Court held that Genentech should not be "shut out from arguing plausibility… simply because they could have pleaded this case in the [parent patent] action and left it too late to seek to amend to include that argument". Summary judgment was refused in respect of those claims as it did not amount to an abuse of process or issue estoppel.