Patents & Designs in the age of generative AI: the bar for obviousness

A legal system designed to reward human creativity

With an underlying purpose of incentivizing innovation by rewarding human creativity, IP rights are monopoly rights, giving the owner the right to stop others using what they have protected. In exchange for a monopoly, the applicant has to share the creation with the world, making a contribution to human knowledge. IP rights can thus motivate further innovation and advance human knowledge. Today’s IP laws were written by humans, for humans, to reward humans – so how can a legal system designed to recognise and reward human creativity be applied in the age of generative AI?

Deriving the right to IP ownership

In most jurisdictions, the right to own a patent or registered design belongs to the creator (i.e. the inventor or designer), or someone who derives that right from the creator. Whilst AI currently lacks the legal ability to own property so lacks capacity to own an IP right anyway, current IP law generally also does not allow for non-humans to be legally named as inventors or designers on patent and registered design applications. It is thus currently not possible to formally recognise in name the contribution of AI to the conception of an invention or design, or to reward the AI system with a monopoly therefor.

This sparks an interesting ethical debate, largely stemming from the matter of IP ownership, which serves the underlying purpose of incentivizing innovation and rewarding human creativity. But even though an AI system cannot be formally named on paper as an inventor or designer and hence cannot be legally recognised as such, generative AI is being used as a tool in the invention and design process today and so may inevitably contribute to the conception of an invention or design.

Ownership and patentability in the age of AI

Entitlement – that’s whether a patent or registered design has been applied for by/granted to the correct owner – is a ground for revocation or rejection in many jurisdictions – so recognising and naming the correct inventors and designers and determining the correct derivation of ownership is important to avoid a loss of rights. But entitlement has no bearing on whether the subject-matter of a patent or design application is considered to be allowable for grant. That is, regardless of who is recognised in name on a patent or design application as its creator(s), the identity of the true creator(s) of the inventor or design has no bearing per se on whether the subject-matter of an invention or design is considered worthy of a patent or registered design being granted.

So even if the contribution of generative AI cannot be formally recognised or rewarded in name as such, how might the actual involvement of AI in the invention and design process be taken into account in the assessment of patent and design applications?

Requirements for registrability

Patents require novelty and inventive step – meaning that to be granted a patent, an invention has to be new compared with what’s been done before, and the new part of the invention should provide some kind of technical advantage, solve a problem, and it shouldn’t have been obvious to modify the pre-existing technology to have that new part. Similarly, registered designs require novelty and individual character – meaning that a design must be new and give a different overall impression from earlier designs.

Setting the bar for obviousness, the inventive step and individual character requirements serve to prevent monopolies being granted for minor and trivial changes. The objective legal tests for assessing inventive step and individual character were written to be assessed from the perspective of legally fictional people who are humans: the “skilled person” for inventive step, and the “informed user” for individual character. The bars set by these tests, as viewed through the lens of the relevant fictional person, define the line between protectable and unprotectable subject-matter, and were written before the advent of generative AI in its current state.

So how might these bars need to be redefined given the contribution that may be made to the conception of an invention or design by generative AI? Where humans and generative AI work together to invent and design, might the current legal standards for assessing novelty, inventive step and individual character no longer be appropriate?

Generative AI and patent law: the skilled person, with the added skill of generative AI

Inventive step is assessed from the perspective of a fictional legal person called the “skilled person”. Case law defines the skilled person as a workman, technician or nerd, who has average knowledge and ability and is not particularly inventive.

Whilst a generative AI system cannot currently be legally recognised in name as an inventor on a patent application, perhaps its contribution to the conception of an invention may nevertheless need to be considered when assessing inventive step. Taking into account inventions created by human inventors with the assistance of generative AI, for example where using AI as a tool or assistant, will the definition of the skilled person need to change?

With the use of generative AI as a tool, engineers and scientists can have a vast amount of knowledge at their fingertips in a short space of time, so in reality, will the skilled person still be of average knowledge and ability? It will be interesting to see whether the definition of the skilled person and hence also the test for assessing inventive step may need to be redefined in future to account for the advantages that AI brings to human inventors.

Generative AI and patent law: generative AI as the skilled person

And what about inventions apparently conceived completely independently and autonomously by generative AI without significant human input, as alleged in the case of Dr Stephen Thaler’s AI creativity machine, DABUS?

Even if such generative AI systems cannot and still will not be able to be validly recognised in name as an inventor on a patent application, will their inventive contribution need to be acknowledged by reconfiguring the definition of the skilled person and the standard for inventive step?

If the legal notion of the skilled person were to be interpreted to encompass non-humans such as generative AI, perhaps the definition of the skilled person would need to be modified to account for the differences between humans and generative AI. On the one hand, arguably generative AI does not have “average knowledge ability” in the same way as a human, since it can access and process vast amounts of information more quickly, and thus has advantages over a human “skilled person”. On the other hand, arguably humans have advantages over generative AI in other ways, because for example, a human can build physical prototypes, do laboratory experiments, and experience real life human problems first hand.

Raising the bar: the impact of AI on inventive step standards

In any case, it seems that what’s obvious to generative AI would not necessarily be the same as what’s obvious to a human inventor. Obviousness, being at the heart of how inventive step is currently assessed, is thus seemingly not the same for humans working with and without generative AI. Regardless of whether legal changes are brought about in future allowing for non-humans such as generative AI to be formally named as inventors, it will be interesting to see how the definition of the skilled person and the bar for inventive step may nevertheless need to be reconfigured.

Designs: the informed user and image based generative AI

It will similarly be interesting to see how the legal test for assessing individual character for registered designs may need to be modified in future. Again, the process of conceptualising and visualising a design differs for humans and generative AI, and with generative AI at the fingertips of designers, new processes and time savings for design ideation and image generation are available.

The assessment of individual character for designs is made from the perspective of the “informed user”, another fictional legal person. A design has individual character if the overall impression it produces on the informed user differs from the overall impression produced on the informed user by the prior art. The informed user is said to have some awareness of existing designs in the sector concerned.

So what about image based generative AI models, trained by the input of pre-existing images? Such models may have been trained with many more design images than what a human “informed user” would ever have seen. To account for the use of image based generative AI by human designers, might this require a reconfiguration of the definition of the informed user or the way in which individual character is assessed?

Conclusion: AI and the future of IP protection

Current IP law generally does not allow for an AI system to be recognised as an inventor or designer. Regardless of whether IP law will be modified in future to allow AI systems to be legally named in this way, it will be interesting to see whether case law might evolve to take into account the contribution of generative AI in assisting humans in the invention and design process. The current tests for objectively assessing inventive step and individual character are viewed through the lens of fictional legal human people, and perhaps these tests will become outdated and no longer appropriate given the use of AI by humans and the advantages that this brings in the conception of new ideas and images. With the advantages of generative AI at their fingertips, perhaps the standard of what is deemed obvious to human inventors and designers may need to evolve, reconfiguring the bar for what’s considered worthy of patent and design protection.