The Court of Appeal in London has ruled against the physicist Stephen Thaler’s latest bid to list his AI machine Dabus as an inventor on a patent.

In what has become a test case for patent law, Thaler has applied for patents listing Dabus as the inventor in a number of jurisdictions, including the US, Europe and China as well as the UK.

Dabus is a so-called “creativity engine” that uses artificial neural networks to generate and assess new ideas. Thaler argues it’s the sole inventor of, among other things, a food container that improves grip and heat transfer, an area in which he says he has no expertise and so couldn’t have contributed to the patent in a way that would qualify him for inventor’s rights.

The High Court in England and Wales sided with the UK Intellectual Property Office last September in refusing the patent applications, accepting that while Dabus created the inventions, it couldn’t be granted a patent on the grounds that it wasn’t a ‘natural person’.

Now, the Court of Appeal in London has found that only a ‘person’ with legal personality can be an inventor, and that as Thaler accepts that he isn’t the inventor, he isn’t entitled to the patent.

The decision puts the UK at odds with South Africa, which in July became the first country to award a patent naming an AI as its inventor and the AI’s owner as the patent's owner when it accepted Thaler’s application.

This was followed a few days later by the Federal Court of Australia making the first judicial determination that an AI can be named as the inventor on a patent.

However, there was disagreement among the Court of Appeal's judges over the requirement for a person applying for a patent who isn’t the sole inventor to name who they believe is the inventor. Lord Justice Birss considered that Thaler had met the requirement by listing Dabus as it was his “genuine belief” that Dabus was the inventor. But Lord Justice Arnold and Lady Justice Laing disagreed, with LJ Arnold saying it is implicit in the wording of the Patents Act 1977 that only people can be inventors and that Thaler’s answer was “a legal impossibility.”

University of Surrey professor Ryan Abbott, who along with his team represented Thaler, said: “We were very encouraged by the dissent of LJ Birss who agreed with us that ‘the creator of the inventions in this case was a machine is no impediment to patents being granted to this applicant...' and he would have allowed the appeal. We respectfully believe this was the correct interpretation of the current state of UK law.”

Sarah Taylor, an IP litigation specialist at Pinsent Masons, commented: “Lord Justice Birss’ dissenting judgment is interesting. He found that the simple fact that this situation had not arisen before should not prevent the court from deciding how the law should be applied in a new situation, perhaps indicating that in some circumstances all that is needed is for the law to be applied in a more agile manner.”

But she added that the ongoing debate and disagreement amongst the leading patent judges makes it likely the matter will reach Supreme Court. “Ultimately the Court of Appeal’s decision has not changed much, and it is still likely that, in order for a machine to be named as an inventor of a patent, legislative change is going to be needed.”

Thaler has been at odds with patent offices around the world for years over the need to recognise AI systems as inventors and has seen his applications rejected by the European Patent Office and the US Patent and Trademark Office.

Abbott agrees, saying that the status quo is no longer fit for purpose and could put investment in AI at risk. Speaking to The Times, he said that naming the creator of the AI system as the inventor was legally risky, as they would not have substantially contributed to what the AI had created and would therefore be breaking patent law.

He also pointed to the increasing use of AI in R&D to discover new drug compounds and repurpose drugs, where there may be an invention that qualifies for a patent but not a person who qualifies as an inventor. If that means a patent won’t be granted, he says, “it signals to companies that are investing in AI, like DeepMind or Siemens or Novartis, you can’t use AI in these areas”.