The Thaler decision – AI Need Not Apply … At least Not yet….

The UK Supreme Court held, on 20 December 2023, that an AI system could NOT be an inventor for two UK Patent applications.

The reasons given are somewhat more prosaic than you might have guessed, except, as always, there are potentially wider implications that were not addressed by the Court.

It wasn’t that big of a deal…

There were no earth-shattering philosophical debates about whether AI should be recognised or endowed with legal personality, nor were there metaphysical musings about what it meant to be living in an age where AI could well create inventions. In the end, it was just a cold hard look at what the UK Patents Act said an inventor is and what the Act intended an inventor to be, and then a logical decision was made, and an AI system just couldn’t fit the bill.

Except it might be…

But that’s only part of the story. And whilst it’s already making its rounds across the IP practice-verse, few commentators, have commented on the wider implications of obviousness in the age of AI, and the use of AI by inventors in the future of inventing.

If you consider that old standard legal chestnut in patent law of the “person skilled in the art” (or PSA), might we now need to take into account the potential that AI augmentation brings to the average PSA? In patent-speak, does the benchmark for the adequate spark of ingenuity now need to be raised, and will obviousness now be a more challenging barrier to obtaining a patent?

 AI need not apply

But let’s first recap what Thaler decided. The Court unanimously dismissed an Appeal by the American computer scientist and applicant, Dr Stephen Thaler. He lost his bid to have the inventor of his two UK patent applications recorded as DABUS, an autonomous AI system that he owned.

The UK judgement, however, was not concerned with:

1. 1. who the inventor was;
   2. the broader question of whether technical advances generated by machines acting autonomously and powered by AI should be patentable; or
   3. the question of whether the meaning of the term “inventor” ought to be expanded, (so far as necessary), to include machines powered by AI.

Instead, the main issue that was decided, was whether the meaning of the term ‘inventor’ extended to a machine such as DABUS, given that Dr Thaler had always argued that the technical advances and new products disclosed in his patent applications were devised by DABUS and that DABUS was the inventor[1].

This is where it got prosaic. The UK Supreme Court decided that from the wording contained within the UK Patents Act 1977, only one interpretation of the term ‘inventor’ was permissible, that it had to be a natural person. It also found that there was no suggestion in any of the Act’s provisions, that an inventor may be a machine and that confirmation of the correctness of this interpretation was to be found in Rhone-Poulenc Rorer International Holdings Inc v Yeda Research and Development Co Ltd [2007] UKHL 43, mentioned in the Appeal Decision.

This is important in Singapore because the Singapore Patents Act very closely resembles that of the UK Patents Act. While Singapore is not bound by UK decisions, it is often guided by them, particularly in areas where it has no case law of its own on a particular point of law.

The UK Supreme Court also decided that Dr Thaler was not entitled to apply for patent protection in respect of any technical advance made by DABUS, as he did not satisfy any part of Section 7 which provides a complete code for who has the right to apply for and obtain a patent, (which to all intents and purposes, is mirrored by the wording in Section 19 of the Singapore Patent’s Act). This was because it did not confer on any person a right to obtain a patent generated autonomously by a machine, let alone someone claiming that right solely because they owned the machine.

As Dr Thaler also failed to file a statement identifying a natural person to be the inventor and explain how he derived his right to be granted the invention from the inventor for each application, both applications were refused.

The main ruling that the inventor must be a natural person follows similar decisions by the EPO and other courts in Germany, Australia, Israel, New Zealand, South Korea, and the USA thus far, providing some level of certainty that an inventor cannot be a machine.

Looking closer at the use of AI in invention

But let’s look closer at Dr Thaler’s case.

There are 2 interrelated angles to consider:

• 1. Whether the gradual availability and prevalence of AI tools and techniques should change the way we think of whether an invention was “obvious”; and
  2. Whether inventiveness may also lie in the use of AI.
  3. To start with the first, a little background is helpful.

In patent law, an invention may not meet the requirements for a patent if it was “obvious”. What is “obvious” has always been judged by a PSA. The PSA is supposed to be a person of ordinary skill in the art, who was well acquainted with “workshop techniques”, had carefully read the relevant literature, possessed an unlimited capacity to assimilate the contents of scores of specifications and other forms of prior art, but was completely unimaginative and incapable of a “scintilla of invention” (that is an actual quote from case law, by the way).

 If, however, the notional PSA is now to be endowed with (or considered to have) access to AI, will this have an impact over what level of knowledge and skill the PSA is to be deemed to have?  After all, even if the PSA is not to be capable of a “scintilla of invention”, might he or she be deemed to at least be able to deploy AI in research?

Another way to put it is to ask: should AI be deemed to be just another “workshop technique” at his / her disposal?

For anyone who has done work with AI, “using” AI is not as “simple” as it might appear. To take generative AI as an example, prompt engineering may well be both a matter of intellectual but unimaginative labour but it can also require instances of inspired insights in working with data to produce effective results.

How would we assess what is “obvious” then? Would it be a question of what series of prompts or data was used to influence output?

And if we agree that the PSA should be augmented with AI, might this all lead to us to revise what is “obvious” and what is an “inventive step”?

AI augmentation: inventors

On the second point, whether inventiveness may also lie in the use of AI[2], to first state the obvious (pardon the pun) the Thaler decision does not preclude a natural person from employing AI to devise an invention, provided that the natural person is ultimately identified as the inventor.

An inventor is, to use a dictionary definition, a “deviser” who plans (i.e. a complex procedure, system, or mechanism) by careful thought or creates or brings about (an object or a situation) by the deliberate use of skill and artifice. What skill sets might an inventor need to use AI to develop his or her invention? What implications does this have for inventors that use AI?

 In the age of accessible AI, there may be nuanced questions that arise, on a case-by-case basis, even as to whether the “inventive step” that elevates an invention from the “obvious” might well lie in the way in which AI was deployed as a tool or in the inventive process.

 In Singapore, like the UK, the inventor is the actual “deviser” of the invention, and unlike for example copyright, where doing little else, other than depressing the shutter of a camera provides copyright protection for an original work, the deviser of an invention needs to do more. (Admittedly, we are deliberately downplaying the skill and intellectual effort to compose the photograph, etc).

The AI “Genie” is out of the bottle

With the AI genie now well and truly out of the bottle, our world is irrevocably changing, and it is clear, that aspects of the current UK and Singapore IP legislation (at least) appear to be increasingly out of touch with this new and rapidly changing reality.

How the law now develops to protect inventions generated with the aid of generative AI powered machines remains to be seen, and the remaining questions of whether:

• 1. technical advances generated by machines acting autonomously and powered by AI should be patentable;
  2. the meaning of the term inventor ought to be expanded to include such machines; and
  3. the PSA will continue to remain as a person of ordinary skill or not;

will inevitably need to be addressed, as the growth of industries that rely on AI in the development of new technologies expands.

What do would-be patent owners need to do?

To conclude, it’s worth observing that some forms of innovation these days are driven by the use of AI. This is a trend that will only accelerate and proliferate.

If you’re using AI in any way to augment your processes to arrive at your invention or if it is integrated as part of your invention in any way, it’s worth seeking advice on how best to do this. Taking a sound legal and technical approach to formulating your claims and thinking through how to structure your patent is an important step.

The machines have not risen. And AI systems need not apply for patents. Yet. But they certainly are among us.

And they’re not going away.

Authors: André Chissel and Jeffrey Lim

Footnotes

[1] An observation, to be discussed perhaps in a later article, is the fact that Dr Thaler thought that DABUS was more than a PSA, being inventive.

[2] A query to ask on this point is whether the use per se of a workshop tool (because it is obvious to try to use the tool) that is as inventive as DABUS lead to an invention that requires a natural person to be the inventor in the first place? A subject for comment and further consideration in another article.