Can an invention enabled by artificial intelligence be patented?

In 2020, the US Patent and Trademark Office (USPTO) denied a patent petition filed on behalf of DABUS, an artificial intelligence (AI) system.

The items DABUS sought patents for were a food container with a fractal surface that helps with insulation and stacking, and a flashing light for attracting attention in emergencies. They were filed by Stephen Thaler, a physicist and the founder of Imagination Engines, a company that researches and develops artificial neural networks like DABUS.

Thaler petitioned that DABUS was a “creativity machine” that recognised the “novelty and salience of the instant invention.” However, in its final opinion, the USPTO ruled that “patent laws require that an inventor be a natural person,” citing previous federal rulings on the language surrounding the nature of invention.

Then, earlier this month, in a parallel case involving a copyright issue with Thaler’s AI system, a US federal circuit court upheld a 2021 decision confirming that, as per the language of the Patent Act, AI systems cannot patent inventions because they are not human beings.

Thaler plans to appeal the circuit court’s ruling, with his attorney criticising the court’s “narrow and textualist approach” to the Patent Act.

The DABUS case fundamentally alters how intellectual property (IP) is perceived, and it reveals what patent systems are starting to grapple with: If AI is responsible for an invention, can the machine be awarded a patent?

“The DABUS case highlights how AI is going to increasingly challenge intellectual property law. AI will be used more and more to help invent,” said Toby Walsh, a Laureate Fellow and Scientia Professor of AI at UNSW Sydney.

“IP law has always struggled to keep pace with technology. Previously it had to deal with the challenges that genetic engineering threw up. It now needs to deal with the challenge that AI throws up,” he told TRT World.

The DABUS case is the first time an AI system has been designated as the sole inventor – a quandary facing lawmakers around the world as they seek to consolidate international legal opinion around AI patent law.

Patents related to DABUS were also denied in the UK, Europe and Australia on similar grounds related to personhood, with the European Patent Office questioning who would enforce the rights granted to an inventor under such a circumstance.

Keeping up with AI

While the DABUS case marks the first time an AI system has been designated as an inventor, it isn’t the first time that AI has been instrumental in the process of innovation.

In 2019, an antibiotic called Halicin was created using a deep learning algorithm that helped identify a chemical compound effective against drug-resistant strains of bacteria. It was drawn from a pool of over 100 million molecules that fight against various pathogens.

“Halicin was originally meant to treat diabetes, but its effectiveness as an antibiotic was only discovered by AI that was directed to examine a vast catalogue of drugs that could be repurposed as antibiotics. So there’s a mixture of human and machine coming into this discovery,” Professor Walsh said in a press release.

In the case of DABUS, it is unclear whether the system is truly responsible for the invention. Professor Walsh highlighted some of the challenges with ownership regarding AI-related inventions.

“There are a bunch of fundamental problems. Who owns the IP? It can’t be the AI. Does AI change the nature of invention? Currently, an invention must not be obvious to an expert capable in the state of the art,” Professor Walsh told TRT World.

For an invention to be patented, it must meet a certain number of requirements, as standardised by the WTO’s 1994 Agreement on Trade-Related Aspects of Intellectual Property Rights (TRIPS).

The first condition is that the invention must be novel, or not already in existence. Then, it must be non-obvious to a person who has expertise in the specific field of invention. Finally, it must be capable of industrial application or utility.

“But an AI might be much more capable. Does this raise the bar for what is obvious? And will patent offices be able to keep up with the speed that AI can invent?” Professor Walsh notes.

“Even if we do accept that an AI system is the true inventor, the first big problem is ownership. How do you work out who the owner is? An owner needs to be a legal person, and an AI is not recognised as a legal person,” said IP law specialist Associate Professor Alexandra George.

In an article for the science journal Nature, co-authors Professor George and Professor Walsh argue that governing bodies around the world will need comprehensive legal reform to determine whether AI systems can be awarded IP protection, proposing the design of a bespoke form of IP law called AI-IP that would be tailored to protect AI inventiveness.

Rather than retrofitting old patent laws to accommodate new technologies like AI, a “distinct AI-IP doctrine has the advantage that it could be tailored to meet the specific conditions in which AI creativity occurs,” they wrote.

“[Whereas] patents are typically awarded to the inventor, lawmakers could decide to distribute the rewards from an AI-generated invention differently – perhaps between the AI developer, the person directing the AI and the owner of the data used to train it.”

They suggest that an international treaty governing AI patents will be essential, much like what the TRIPS agreement did in governing the use of designs, trademarks, copyrights and other areas of IP. “It would set out uniform principles to protect AI-generated inventions in multiple jurisdictions,” they added.

But if a legal framework for patenting AI-generated inventions is not achieved, there will be implications in attracting investment in cutting-edge industries.

“Funders and businesses would be less incentivised to pursue useful research using AI inventors when a return on their investment could be limited,” they said.

“Society could miss out on the development of worthwhile and life-saving inventions.”