March 25, 2021

Are AI-generated inventions patentable?

Products created by AI programmes are becoming common, but outdated laws mean they have little protection from copycats.

Andreas Leupold

7 min read

What does a chair from furniture manufacturer Kartell have in common with a rocket engine by the software powerhouse Hyperganic? They were both created by generative design — in other words, made by AI.

Our patent and copyright system isn't prepared for inventions or designs by AI.

The buzzword 'generative design' stands for a computer-aided design (CAD) process. But it's a far cry from simple CAD design, using algorithms created by AI to generate a first set of designs for a product based on certain input parameters. It will then continue to refine these designs with each iteration until the final product materialises.

Combined with industrial 3D printing, the result is a technically superior product that weighs less, has better functional features and is often less prone to wear and tear.

A rocket engine designed by Hyperganic.

And this combination is discovered by more and more industry heavyweights: Airbus is using generative design and 3D printing for aeroplane interiors; US sports equipment company Under Armour is using it for improving the damping properties of its 3D printed running shoes; and German car manufacturer BMW is using it to create automotive parts that are up to 50% lighter than their conventional counterparts.

But generative AI design won’t just disrupt conventional design thinking. It will also change our patent and copyright system, which isn't prepared for inventions or designs by AI.

Creative machines lose the first legal round

As chance would have it, the same year — 2019 — that Kartell presented its AI chair at Milan's furniture fair, Stephen Thaler, a pioneer in 'artificial invention' claimed that he had triggered “the big bang of machine intelligence with his 'creativity machine', the 'DABUS'." Dabus was hailed by its creator as the first “true artificial inventor“ — not a physical machine, but a concept that employed artificial neural networks and parallel computing to generate new ideas or even create art.

He applied for patent protection in the US, the UK and Europe based on his ownership of said machine — not for the machine itself, but for the inventions it had made. The patent offices of all three jurisdictions denied Thaler’s petitions on the grounds that only natural persons, not machines, can be inventors and that patent law consequently does not permit a machine like DABUS to be named as the inventor on a patent application.

For the legal community it was a wake-up call that a paradigm shift on inventorship was on its way.

For Thaler this was a defeat. But for the legal community it was a wake-up call that a paradigm shift on inventorship as we know it was on its way and had to be dealt with.

The World Intellectual Property Organisation initiated an ongoing discussion on intellectual property and AI that continued in 2020. It led to well over 80 questions on how the law should deal with the challenges to industrial property rights and copyrights posed by AI.

Under Armour 3D printed shoe
Generative design and 3D printing were used to create the lattice sole of this Under Armour shoe.

While it is virtually undisputed that AI inventions are not patentable under the current law in the EU, UK and US, with the number of works and inventions being created by AI increasing, there is growing momentum around the idea that they need some kind of legal protection.

Just what the right vehicle might be is still controversial, however. Patent protection was initially devised to help motivate human inventors to innovate and share the results. But unlike human inventors, AI cannot be motivated to innovate by the prospect of obtaining patent protection but only be instructed to innovate regardless of patent protection.

What about copyright protection?

What about copyright protection then, you may ask?

The UK Copyright, Designs and Patent Act (CDPA) states that in the case of computer-generated works, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken.

“Computer-generated”, in relation to a work, means that the work is generated by computer, in circumstances such that there is no human author of the work. It sounds like a promising solution to the AI creation problem, but it isn’t.

It sounds like a promising solution to the AI creation problem, but it isn’t.

The scarce case law on these provisions determines the “person who made the arrangements necessary” to be the person who devised the appearance of the various elements of the work in question and the rules and logic by which each element was generated.

This view seems still to be based on the notion that a computer is only a tool that is used by humans to create works of authorship.

If an algorithm created by machine learning generates a design for a bike frame, the appearance of its elements isn't devised by a person, but the algorithm. In this case, it is hence ill-founded to treat the algorithm the same way as a simple pen and “to suggest that, if you write your work with a pen, it is the pen which is the author of the work rather than the person who drives the pen” as a British court held in the case Express Newspapers plc v. Liverpool Daily Post & Echo in 1985.

Sifted's Deeptech Newsletter
We're planning the launch of a newsletter focused on deeptech. Want to get involved? <a href="https://mailchi.mp/sifted/deeptech-newsletter">Sign up here</a> to help us beta test the idea.

A pen is just an inanimate object that does not create anything by itself, so it is indeed just a tool in the hands of the artist. By contrast, AI is much more than just a tool as it determines autonomously how to shape a product design and then adapts the final version to key learnings from its practical use.

Besides, the British concept of computer-generated works doesn't fit with the European principle of creation and the established case law of the European Court of Justice that only human creations are protected by copyright.

Even so, businesses will have no motivation to make costly investments in AI, if the work results of AI could be used and copied freely by anyone, who would rather save substantial R&D costs and present other people's products as their own.

We are not talking about a mere theoretical risk here — our legal systems wouldn't afford inventions patent protection and protect product designs, if failing such protection there wouldn't be a tangible risk of exploitation by free riders. As AI becomes available to such free riders as well, it will make it even easier to create knockoffs from successful designs so we will likely experience a surge of such activities.

So how could protection work?

Which brings us to the shape of things to come — not if, but how will AI-generated works be protected by law in the future?

The best way to provide adequate protection to such works would be the creation of a unique right, similar to the exclusive right of the maker of a database to reproduce and distribute the database in whole, or qualitatively or quantitatively substantial parts thereof.

The best way would be the creation of a unique right, similar to the exclusive right of the maker of a database.

Such a one-of-a-kind right should be afforded to both the algorithm and the designs it generates. This right should not require an invention or the creation of a work of authorship by a human, but only a sizeable investment in the software that created the algorithms that determine the shape and other properties of the final product they create. The owner of such a unique right will be the individual or legal entity that made the investment — only the investor will have the exclusive right to tweak the algorithm and/or use and exploit a product design generated by the algorithm.

Since the law should not stifle progress and economic growth, there is no denying that we need to let go of the idea that only works created by humans are worthy of protection and get to work on a unique right that rewards investments in future-proof AI.

HAL 9000 from 2001: A Space Odyssey would have loved this.

Until legislators catch up, protecting the algorithms as business secrets is the key for recouping investments in AI. If all legal requirements are met for every market, the business secret holder has an arsenal of rights and claims that he can raise against any infringer who uses his proprietary algorithms for making infringing products. HAL 9000 from 2001: A Space Odyssey would have loved this.