It would be too risky for intellectual property offices to accept the notion of a legal ‘e-person’ when assessing inventions generated by artificial intelligence, in-house counsel have said, adding that it is “beyond imagination” at this stage.
Counsel at Continental, ZF Group and BSH Home Appliances discussed the existing system for assessing AI inventions in Europe and China, and whether any changes to the law or examination guidelines would improve the landscape for patenting AI inventions.
Xuting Huang, patent counsel at automotive parts maker ZF Group in Germany, said there is “no doubt” that AI is capable of inventing but that the question of whether it can be legally regarded as an inventor – at the moment the status quo is that it cannot – hinges on whether it can be considered a legal ‘person’.
“There is a risk in accepting a concept of an AI legal person,” Huang explained. “The extreme diversity of AI-based products will in turn lead to an extreme diversity of ‘e-personalities’. It would be very difficult to harmonise these concepts into one definition of an ‘AI legal person’.
“How this can be implemented legally is beyond my imagination at the moment,” Huang added.
Huang was speaking during a panel session on legal technology and AI during Managing IP’s virtual European Patent Forum on Tuesday, September 15.
He was joined by Andreas Reuter, senior IP counsel at automotive supplier Continental, and Thomas Buchholz, chief digital patent counsel at BSH Home Appliances, both also based in Germany.
Buchholz at BHS – the umbrella company of brands including Bosch – said he is “not completely convinced” whether AI can be an inventor.
DABUS debate
He referenced the DABUS cases, in which human applicant Stephen Thaler designated an AI machine called ‘DABUS’ (Device for the Autonomous Bootstrapping of Unified Sentience) as the inventor of a patent application. His applications, filed at the EPO, USPTO and UKIPO, have all been rejected.
“At present, the person who provides the ‘seed concept’ to solve a specific problem, operates the program, and interprets the output should be named as the inventor,” Buchholz said.
However, the situation may be far more difficult in practice, he added, as there are many different people involved in the implementation and development of an AI-generated invention, making it difficult to determine any one ‘person’.
“In Europe, determining inventorship is not so much of a problem. Whoever submits an application can claim ‘X is the inventor and only someone who has better rights can expel this person from inventorship’,” Buchholz added.
In the US the situation is more complex, he said: inventorship can be challenged in post-grant reviews, via an invalidity defence or by a claim of inequitable conduct.
Eye on China
In China, there are high numbers of AI-related patent applications, said Huang at ZF Group. In 2019, there were 1.4 million patent applications filed at CNIPA, the IP office, almost 10% of which were for AI-related inventions.
Although eight of the top 10 applicants were domestic companies (such as Baidu and Alibaba), Microsoft and Samsung were also among the top applicants, Huang noted.
Huang added that changes to patent examination guidelines, including a new rule allowing claims that contain technical features to be assessed and not excluded outright, were brought into force at the turn of the year.
The same is true of algorithm features, Huang said. The technicality of such features is established when they relate to a technical problem that must be solved, in particular, when data processed via an algorithm is associated with real technical meaning and the algorithm directly solves the technical problem, he explained.
Law playing catch up
Earlier in September, and prompted by a consultation on AI and IP published by the UKIPO, Managing IP asked if the law – already some way behind AI – can ever keep pace with rapid technological advancements.
Reacting to this question, panellists said the law should be continually reviewed but that immediate changes may not be necessary.
Reuter at Continental said the law must be solid but should not react to changes too early.
What is worth noting, however, is the “enormous speed” at which new AI publications are released, Reuter said.
“In the 18 months between filing a patent application and publication of that invention there will be a ton of new material created and the patent might become difficult to enforce. You would need to look back at what was happening at the time of the application, which may lead to problems,” said Reuter.
People skills
Huang at ZF Group suggested that the main point to address immediately is the definition of “a person skilled in the art” in relation to patent applications.
“This may need to be adapted in order to account for the different technologies and capabilities of AI, which are far beyond a skilled person’s (human’s) capability.”
As long as a skilled person remains imaginary, though still a “natural” person”, there will unlikely be changes to this assessment, Huang predicted.
Buchholz noted that when there are many possible solutions at play and an inventor finds a solution it is fairly straightforward to claim inventive step successfully.
“By using AI, it is much easier to try millions of outcomes and evaluate them. Perhaps this will result in the argument for inventive step becoming less convincing in future,” he suggested.
