The year 2021 has seen notable developments in AI-related patent applications and open source software licensing. However, unlike the past few years, 2021 has been a relatively quiet year in the law on material eligibility, i.e. what can be patented. Here, we review recent developments regarding: 1) subject eligibility for computer-implemented inventions; 2) AI-related patent applications; and 3) free software licensing.
Subject eligibility for computer-implemented inventions
In Canada, 2021 was marked by the absence of significant new developments regarding the admissibility of materials. However, developments in 2019 and 2020 continue to provide patent applicants and patent examiners with more clarity regarding subject matter eligibility, particularly for computer-implemented inventions. Recent developments have tended to bring the Canadian approach to determining material admissibility closer to the American approach, but with some elements of the European approach as well.
To be patentable in Canada, an invention must meet four main criteria: it must be new, it cannot be obvious, it must be useful and, above all, it must meet the definition of “invention” (as defined in the patent law). On this last point, CIPO has, in recent years (from 2011 to 2020), used a “problem-solution” approach to the interpretation of claims when determining the admissibility of the subject matter of inventions. computer-implemented (for example, inventions of software or business methods). This has led patent examiners to often characterize the alleged “problem” in such a way that all computing elements cited in a claim were deemed “non-essential” to solving the problem and therefore ignored. The remaining “essential” claim elements would then easily be classified as an abstract idea or a simple diagram and thus the entire invention would be excluded from patentability.
The “problem-solution” approach was in some ways similar to the European Patent Office’s (EPO) “technology contribution” approach to patentability. During the EPO’s inventive step analysis, the prior art is compared to the claimed subject matter in order to distinguish novel features from those of the closest prior art. Only those features which contribute to a technical solution (i.e. an “additional technical effect”) of the objective technical problem (i.e. the problem solved in relation to the state of the art most nearby) are taken into account. Any non-technical feature of the invention is effectively considered to be prior art for the purposes of assessing inventive step.
In 2020, the Federal Court of Canada (FCC) ruled1that the Commissioner of Patents had erred in using the problem-solution approach, rather than the “purposive construction” approach set out by the Supreme Court of Canada.2 The “purposive construction” approach takes into account the intention of the inventor and the terms of the claim to determine whether particular elements of the claim should be considered essential or non-essential to the invention.
In response to Choueifaty, CIPO has published a Notice of Practice – a guidance document for patent examiners – to assist in the proper application of purposive construction. These guidelines are now followed by patent examiners and have caused examiners to stop using the “problem-solution” approach to claiming interpretation and instead apply the purposive interpretation test. Another peculiarity of the CIPO guidelines is that they suggest a “real invention” test in which an element is “not part of the real invention” if it does not cooperate with other elements of the claimed invention” (emphasis added).
This new approach resembles the two-step subject eligibility test that is used in the United States. If the claims are found to relate to an abstract idea during Step 2A of the US test, examiners are asked to determine whether the abstract idea is sufficiently “integrated into practical application” and additional elements – such than a computer – must be meaningful limitations that provide more than “extra-solution activity”.
The Patent Appeal Board (PAB) of Canada, in its reconsideration of Choueifaty’s application, ruled that in Choueifaty’s invention “the calculations computerized herein are not intended merely to provide information, but to enable the computer to perform portfolio optimization procedures with much less processing and greater speed.” In other words, the computer in Choueifaty’s claims is not just a computer, it is a better computer. More importantly, the computer is better thanks to Choueifaty’s algorithms. Or, as the PAB puts it, “the algorithm improves the functioning of the computer used to run it […] the computer and the algorithm together form a single real invention which has a physical character and solves a problem related to the manual or productive arts.”
This new Canadian approach, similar to the American approach, is likely to favor inventions that serve to improve the functioning of computers, for example by increasing speed, efficiency, security or reliability. However, it is likely that CIPO will continue to reject computer-implemented invention claims where the computer is incidental or tangential to the invention, where the remaining elements of the claim can be classified as mere scientific principles. or abstract theorems. As always, careful and thoughtful patent drafting will continue to be important in crafting appropriate claims and avoiding such potential pitfalls.
AI-related patent applications
The greater clarity provided by recent developments in determining subject eligibility is likely to contribute to the continued increase in patent filings for fundamental AI technologies (e.g., machine learning, neural networks) and the application of AI technologies in sectors such as healthcare, manufacturing and transport. In particular, we have recently highlighted the acceleration of patent activity in applications such as image recognition, genomics and the human microbiome.3
As AI technologies continue to evolve, we can expect the contribution of AI to the invention process to continue to increase. Currently, AI is one of the many tools used by humans to invent. However, there are signs that AI may soon play a bigger role. Therefore, an emerging question is whether an AI machine can be listed as an inventor on a patent and, if not, who is the actual inventor? Recent rulings suggest that it will be some time before AIs can be considered inventors.
The UK Court of Appeal recently dismissed an appeal by Dr Stephen Thaler and upheld decisions by the UK Intellectual Property Office (UKIPO) and the High Court of England and Wales, finding that an inventor must be a natural person and therefore a machine cannot be considered an inventor.4Thaler is the creator of the Device Autonomously Bootstrapping Uniform Sensibility (“DABUS”), a so-called “Creativity Machine” that uses neural networks to generate new inventions. Thaler applied for two UK patents in 2018 naming DABUS as inventor, which UKIPO rejected. Thaler has also filed patent applications with the European Patent Office (EPO) and the United States Patent and Trademark Office (USPTO), with both offices later rejecting the applications. Thaler won a victory in Australia with the Federal Court ruling that an AI machine can be named an inventor under the Patents Act 1990 (Cth).5 However, it seems likely that in most jurisdictions, patent offices and courts will wait for lawmakers to change patent laws if AI machines are to be considered inventors.
Open Source Software Licensing
Computer software permeates all of society. While software development tools have grown in ease of use and abstraction, the software itself is more complicated than ever. It’s not uncommon for even simple applications to use dozens of ready-to-use software libraries and resources. This has been accompanied by an increase in the use of open source software in the development of proprietary commercial computer software, including software embedded in hardware devices such as consumer electronics. The year 2021 has seen notable developments regarding the licensing of open source software and the implications of incorporating open source software into proprietary commercial products.
Some open source software is made available under what are known as “copyleft licenses”, which generally require that the distribution of any software incorporating or interacting with the open source software be governed by the same copyleft license. The additional terms of such a copyleft license may require that source code for proprietary software – including proprietary or commercial software – be made available to the public.
On October 19, 2021, the Software Freedom Conservancy (SFC) sued Vizio, Inc. for alleged violations of the GNU General Public License (GPL) covering software embedded in certain Vizio smart TVs. In its complaint, the SFC alleges that Vizio included the Linux kernel and several related open source programs in its SmartCast system for televisions. SFC further alleges that the sale of these Vizio smart TVs constitutes distribution of the GPL-licensed software, and that Vizio is therefore in breach of its obligations under the GPL license. Specifically, SFC asserts that the SmartCast system is covered under the terms of the GPL and therefore the source code for the SmartCast system must be made available by Vizio.
Vizio’s lawsuit underscores the importance of understanding the requirements and risks of using open source software, and in particular, integrating it into proprietary commercial products. While open source software can be a powerful tool in agile software development, it’s critical that proprietary software developers create strong strategies and policies to understand and adhere to applicable open source licenses.