UK Court of Appeal says AI is not an Inventor and is split on allowing applications for AI inventions

Court at First Instance

Mr Justice Marcus Smith upheld the Hearing Officer’s decision (Thaler v Comptroller General of Patents, Trade Marks and Designs [2020] EWHC 2412 (Pat)). 

Court of Appeal

On appeal, the following 3 issues were considered. 

1. Does an “Inventor” have to be a “Person”?

The Court of Appeal unanimously held that the Patents Act 1977 was drafted on the footing that an “Inventor” had to be a “Person” (examples cited include ss.7(2), 7(3), 8 and 13).  Consequently, the UK patent system currently will not recognise an AI machine as an inventor.

2.  Was Dr Thaler entitled to apply for a patent for an invention created by DABUS?

Dr Thaler accepted that under s.7, he could not apply for the grant of a patent as the inventor (s.7(2)(a)) or as a successor in title (s.7(2)(c)).  So, Dr Thaler relied on s.7(2)(b).  He argued that under the law of accession, as owner of DABUS, he was entitled to the property in any of DABUS’ inventions in the same way that an owner of a tree also owns the fruit produced by that tree.  Dr Thaler’s counsel, however, had to admit that the law of accession had only ever been applied to situations where new tangible property (e.g. the fruit) had been produced from existing tangible property (e.g. the tree).  Furthermore, Lord Justice Arnold saw that there were 2 problems in expanding the law of accession to cover a situation where new intangible property (the invention) was produced from existing tangible property (the AI machine).  First, as Blackstone notes most of the law of property is based on the concept of “exclusive possession” i.e. the owner of the tree has exclusive possession over the tree and the fruit produced from this tree and so is able to own both.  Arnold LJ noted that this does not necessarily apply to intangible property which cannot be exclusively possessed e.g. “consumption” of intangible property by one person does not prevent simultaneous consumption of the same property by another person.  Second, Arnold LJ noted that “an invention is a piece of information” (Merrell Dow) and information cannot be owned as property.  What s.7(2)(b) means by “property” is not the invention or information but the right to apply for an invention.  The tree/fruit analogy under the law of accession could therefore not be applied to the present case with DABUS/invention as the facts were not the same.

Given Birss LJ’s analysis on (3) below, he considered that it was unnecessary to examine whether there was any such rule of law (whether under the law of accession, or otherwise) on which Dr Thaler could rely to make the application for the patent.

3. Did the UK IPO respond correctly to Dr Thaler’s patent application?

On this question, Lord Justice Birss reviewed the legislative history of s.13(2) and the case of Nippon Piston Ring [1987] RPC 120.  From these sources, Birss LJ inferred that the purpose of s.13(2) was to ensure that bona fide errors in identifying the inventor or the derivation of the applicant’s title should not prevent an applicant from obtaining a patent or render the patent granted liable to revocation.  He also held that it was clear that the purpose of s.13(2) was not to enable the Comptroller to investigate the factual correctness of the answers given by the applicant in its application forms.  Arnold and Laing LLJs agreed with Lord Justice Birss’s analysis up to this point – but from this point their analysis differs.

For Birss LJ, because Dr Thaler had honestly tried to identify the information sought by s.13(2) (i.e. the identity of the inventor and the applicant’s derivation of title), Dr Thaler had complied with his obligations under s.13(2).  Birss LJ did not think the Comptroller was entitled under s.13(2) investigate the “factual correctness” of the inventor’s identity or Dr Thaler’s derivation of title.  Importantly, Birss LJ noted, being granted a patent does not ratify that your claim to derivation is good.  If someone thought they had a better claim to that patent than Dr Thaler, they could apply under s.8 (pre-grant) or s.37 (post-grant) to correct the proprietorship register for that patent.

Arnold and Laing LLJ did not agree with Birss LJ that the hearing officer had investigated the factual correctness of the inventor’s identity or Dr Thaler’s derivation of title.  As Arnold LJ noted the hearing officer, equipped with the knowledge of the correct law (as set out in (1) and (2) above), had seen that Dr Thaler’s application was deficient on the face of the form as it had failed to provide the information requested by s.13(2).  Consequently, according to Arnold and Laing LLJ, the hearing officer was justified in deeming Dr Thaler’s application as being withdrawn.

Conclusion

This is not the first time that inventors have tried to fit a new kind of invention into the existing patent framework.  Swiss form claims are an example.  However, this case shows the difficulties of fitting AI inventions within the current legislative framework.  Birss LJ noted that Dr Thaler’s case was “frequently put on the basis of what the law ought to be rather than what it was” [5] Birss LJ’s judgment attempts to provide a pragmatic solution to this issue but leaves us with a “missing inventor” problem, i.e. we would have a patented invention but no inventor since DABUS cannot be recognised as the inventor under the Patents Act 1977 and Dr Thaler did not think he was the inventor.  Birss LJ’s solution achieves a way of including AI inventions within the current statutory framework but inevitably is not a perfect solution given that the Patents Act 1977 was not drafted to accommodate them.  Arnold and Laing LLJ’s judgment, on the other hand, remains more loyal to the wording of the Patents Act 1977 but in so doing prevents the possibility of AI machines being recognised as inventors.

An alternative solution, hinted at by the hearing officer, Mr Justice Marcus Smith and Birss LJ [6] would have been for Dr Thaler to apply as the inventor.  There appears to be a feeling that such an application may have been more successful.  Presumably, this route was not taken because (1) it is contrary to the key selling point of DABUS – that DABUS (rather than Dr Thaler) created the inventions, and (2) there is a risk that if a patent was granted on this basis it could be revoked on the grounds that it was granted to a person who was not the actual deviser of the invention (and had no derivative right from such a person).

It will be interesting to see if this case is appealed to the Supreme Court and whether it triggers a legislative response.  Last year, the IPO launched a consultation seeking views on how artificial intelligence should be accommodated by the UK patent system, replies were received from 92 stakeholders and on 23 March 2021 the Government published its response (please see here).  For now, it appears that practitioners will not have to grapple with the tricky questions that could have followed from the Court of Appeal’s judgment such as, how should we assess inventive step under the Patents Act 1977 when the invention is made by an AI machine.