Does EU Patent law need to change to account for AI?

Karl Barnfather, chairman of intellectual property firm Withers & Rogers, argues the case for EU patent law to adapt to properly cover innovation

Until relatively recently, inventors have been human. However, as machine learning becomes more common, the question of whether European patent law needs to evolve has become more pertinent. Whilst interesting AI innovations have recently secured patent protection, some aspects of European patent law may still need to catch up.

Using Artificial Intelligence (AI) systems to invent is nothing new. In fact, scientists at Dartmouth College embraced it as a tool for conducting scientific research back in 1956. However, the recent breakthrough we have seen in machine learning's success can be attributed to the increase in the amount of available data and access to cheap, scalable computing power.

Due to the rapid expansion of areas such as online communication and the Internet of Things, the amount of data available has grown exponentially, enabling AI and machine learning to gain traction. For example, the Urban Software Institute has recently developed a method for monitoring traffic control systems, described in patent EP3144918B1. To help improve traffic flow, the system includes multiple traffic lights that predict future signal status using a machine-learning algorithm.

While some issues regarding patentability remain, case law indicates that AI inventions are more than capable of securing patent protection. Whereas patent protection for an invention disclosing the use of AI or machine learning to solve a technical problem remains much the same as for any other computer related invention, there are unresolved issues relating to ownership and inventorship when AI or machine learning was used to help "create" the claimed invention.

At present, patent applications must identify one or more human inventors, which to date has not caused any issues due to the element of human input still required. However, in the future, a lack of clarity around the inventorship of AI inventions could become an area of contention.

The fact that such innovation is already in use makes resolving the issue of inventorship even more timely. In the 1980s Stephen Thaler, president of US firm, Imagination Engines Inc, began experimenting with software which modelled the neurological processes of the human brain, named The Creativity Machine. The machine created new ideas and adapted to different scenarios without human intervention.

The research led to the invention of the Oral-B Cross Action toothbrush, and whilst Thaler listed himself as the inventor of the patent, arguably the real inventor was in part The Creativity Machine.

Correctly identifying the inventors of an invention is an important process as it helps determine the ownership of the patent. In most countries, whoever applies for a patent is presumed to be the inventor and owner of the invention and entitled to the grant of the patent.

When this is not the case, the inventor(s) sometimes need to assign ownership rights to the respective proprietor. At present, it is unclear how such an assignment would take place when the inventor, or co-inventor, of an invention is an AI or machine learning system.

Given the increased use of AI systems in research activity, there should be a mounting number of patent applications claiming some form of computational inventorship. However, a lack of understanding about the patentability of software-based inventions and computational systems could be putting innovators off from seeking patent protection.

Crucially, innovators should be reassured that in many cases these inventions are patentable as long as they can demonstrate a technical effect solving a technical problem.

At present, the owner of the machine or AI system is assumed to be the inventor of the patented invention. However, this may need to change in the future and patent law would need to adapt accordingly.

Invention rarely occurs in a vacuum and with companies increasingly involved in open innovation, data-based machines could be used even more widely, to deliver multiple outputs, in the years to come.

This could make it harder to establish who is the rightful owner of any IP generated due to a lack of clarity related to inventorship. For this reason, it is even more important for those involved in such collaborations to put in place a clear, contractual agreement, which is agreed at the outset.

As patent applications of AI inventions increase in number, EU patent law may need to adapt further. While this shouldn't put inventors off from pursuing legal protection, they should bear in mind that in doing so they will be helping to shape this brave new world of innovation.

Karl Barnfather is chairman of intellectual property firm Withers & Rogers. The article was also supported by Dr Harry Strange, Trainee Patent Attorney at the firm.