‘DABUS’: the AI topic that patent lawyers should be monitoring
Imogen Ireland and Jason Lohr of Hogan Lovells examine the recent decisions on “DABUS” to show why AI inventors are not just a future problem
In 2018 and 2019, parallel patent applications were filed at the USPTO, EPO and UKIPO by the applicant – and human – Stephen Thaler. Where an applicant might ordinarily complete the details of the inventor, it was instead explained that said inventor was an artificial intelligence (AI) machine called "DABUS" (short for Device for the Autonomous Bootstrapping of Unified Sentience). DABUS was described as "a type of connectionist artificial intelligence". It was also said that DABUS had "identified the novelty of its own idea before a natural person did" and therefore the machine should be recognised as the inventor.
In decisions handed down in 2019 and 2020, all three offices refused the applications. The primary reason for these outcomes concerned the fact that DABUS was a machine, whereas the legal frameworks applied by the EPO, UKIPO and USPTO require the inventor to be a natural person, or human. What is particularly interesting about these decisions is that, until they were given, there was little guidance – or debate – addressing inventorship of inventions made using AI in the US, UK, or Europe as a whole.
The EPO and UK approaches to inventors
It is worth considering the UK and European law approaches to AI and inventors in parallel. The initial decision by the EPO foreshadowed international regimes, and, as the UK Supreme Court recently reminded us, the "interpretative objective is to strive for consistency between European and UK patent law" ( UKSC 27).
Under the European Patent Convention (EPC), the rules require the patent application to designate the inventor in terms of their family name, given names and full address. In its DABUS decisions, the EPO considered that the interpretation of the legal framework of the EPC led to the conclusion that the inventor designated in a European patent had to be a natural person (and not any other entity). The EPO reasoned that inventorship conferred various legal rights, but that to exercise them the inventor had to have a legal personality, which "AI systems or machines do not enjoy". It also suggested that this requirement for a human inventor was likely to be internationally applicable.
The DABUS decisions raise some important questions about how to handle ‘inventions’ that were at least partially developed or conceived by AI.
In the UK, patent law rules around inventorship are expressed slightly differently. The inventor is said to be the "actual deviser" of the invention. According to case law, there are two requirements underlying this concept. First, they must be a natural person. Second, they must have contributed to the inventive conceptive. The UKIPO's DABUS decision centred on this first requirement and came to similar conclusions to the EPO. Referring to the legislative framework that also underlies the EPC, the UKIPO agreed that a natural person had to refer to a human.
Notable, however, were the UKIPO's concluding remarks. "AI machines are likely to become more prevalent in the future," it said, and although the present system did not cater for such inventions because the technology was simply not around at the time the legislation was being drafted, "there is a legitimate question as to how or whether the patent system should handle such inventions" because "times have changed and technology has moved on". The UKIPO went on to say that this issue should be debated more widely, and noted that changes to the law should be considered in the context of such a debate and not shoehorned arbitrarily into existing legislation.
The DABUS decisions in the EPO and the UK are now under appeal. In the EPO, appeals were submitted in May 2020 and are pending. The decision of the UKIPO was reviewed in the England and Wales High Court by Mr Justice Marcus Smith on July 15 2020 and a decision is awaited. So, we must continue to watch this space.
The US's take on inventors
In the US, the USPTO issued a decision in February 2020, denying a petition to vacate a Notice to File Missing Parts. The application papers, as originally filed, included an application data sheet (ADS) listing DABUS as the sole inventor. The Notice to File Missing Parts indicated that the ADS did not "identify each inventor by his or her legal name." As in the UK and EPO cases, the applicant, Thaler, argued that the invention was conceived by a "creativity machine" named DABUS that should be listed as the inventor.
The USPTO, in denying the petition, pointed to 35 USC 100(f) to clarify that the term "inventor" means the "individual or … individuals collectively who invented or discovered the subject matter of the invention." The USPTO indicated that, to the extent the applicant argues that an inventor could be construed to cover machines, "the patent statutes preclude such a broad interpretation." The USPTO further indicated that "Whoever" suggests a "natural person," citing the dictionary definition of the term, and pointed to the fact that the inventor who executes an oath or declaration must be a "person." The applicant's reading would allegedly "contradict the plain reading of the patent statutes that refer to persons and individuals."
The USPTO also pointed to the University of Utah and Beech Aircraft decisions by the US Court of Appeals for the Federal Circuit where the court indicated that "the patent laws require that an inventor be a natural person." Finally, the USPTO pointed to the Manual of Patent Examining Procedure as indicating that "conception" is defined as "the complete performance of the mental part of the inventive act" formed in the "mind of the inventor." The USPTO then determined that the notice was properly issued because the application had failed to provide the legal name of the human inventor.
In the petition, the applicant had pointed to the EPO and UK DABUS decisions that indicated DABUS had created the invention. The USPTO indicated that it "has not made any determination concerning who or what actually created the invention claimed". It further pointed out that those decisions interpreted "their own respective laws" and that US patent law "does not permit a machine to be named as the inventor in a patent application."
What's the big deal?
The DABUS decisions raise some important questions about how to handle 'inventions' that were at least partially developed or conceived by AI. DABUS was said to have come up with the innovation of its own accord. One of the reasons put forward by Thaler and his legal teams for seeking to recognise DABUS as the inventor was that to do otherwise would mislead the public.
Just extending this idea, one of the benefits of the patent system is that the patent registers and disclosures are public. Commentators on the development and use of AI have expressed a need to monitor the responsible use of AI (Stephen Hawking was one such advocate of this approach). Recognising AI as inventors would encourage this because the disclosure available through a patent register would assist in publicising the way in which AI is being put to use. This of course raises further questions as to whether or not patent law is the correct framework through which to mentor society's development of AI.
With this in mind, it is perhaps not surprising that the UKIPO urged the issue of whether or not to recognise the AI inventor more widely. This issue also illustrates the difficulties the DABUS applications posed the various offices at which they were filed – in many respects, the problem went far beyond the rules on applying for a patent and had the propensity to touch multiple legal regimes.
The problem went far beyond the rules on applying for a patent and had the propensity to touch multiple legal regimes.
That aside, we are still left with the practical matter of being able to obtain a patent. If an invention is discovered by AI, then under the current legal frameworks it is looking like it will be difficult to obtain a patent. Some might see the DABUS example as a 'futuristic model', in other words, exemplifying the use of AI that is, as yet, not widespread in innovation because the state of the art still requires significant human input.
But therein lies the issue – to what extent is the human's input enough to still warrant the title of inventor? This is a difficult question to answer. For instance, in the life sciences industry, various applications (using machine learning) are applied to drug discovery, even to the point that, today, the technology is able to suggest, from many potential candidates, just a handful of lead candidates for clinical study. In this scenario, it still may be possible to identify the human inventor because they have taken the AI's output and done enough further work to show that it can in fact be applied as an invention. However, with AI advancing, it is not unimaginable that in the near future, the human's role will be reduced.
This raises issues around requirements for being an inventor. For example, in the UK the "actual deviser" must not only be human, but also have contributed to the inventive concept. Similar requirements exist in other jurisdictions, such as the US. This point was not clarified in the DABUS applications, and it is arguably the most pressing for the near future. It is conceivable that AI will become sufficiently advanced to the extent that it will be capable of recommending innovations that humans merely have to test or verify. Applying the UK case law to this scenario, for example, the human(s) are unlikely to have met the criteria for inventorship. The result when applying the laws of the US and elsewhere is unclear.
Further, under the DABUS decisions, the AI tool will not be able to obtain the resulting patent. This could result in non-patentable inventions, or non-inventions. It will be interesting to see whether the appeals in the EPO and UK of the DABUS decisions shed any further light on this.
Although the law in various jurisdictions is unlikely to threaten the patentability of innovation arising out of the present-day use of AI, developments in AI are likely to test these limits, perhaps sooner than anticipated. Instead of seeing the DABUS cases as exemplifying a future problem, we should recognise the grey territory that results.