The intellectual property system just doesn’t work for AI

On 10 September we hosted an virtual event on intellectual property and artificial intelligence to explore the questions and concerns that come up. You can access the recording here.

Do patents work for AI?

It was clear from the discussion that AI innovators face a unique set of issues when protecting their data, technology and inventions. This is a fast-moving area – innovations may only have a short effective lifespan. Innovators want, ideally, to protect their patch so that competitors cannot use the same approach. Patents are, in theory, ideal for this purpose, with their solid, monopoly protection. However, the slow-moving and costly patent process, designed with a 20-year time scale in mind, may not be a good fit for AI and machine learning innovations.

Of course, patent applications can be filed, and then only taken forward after the initial year allowed under the Paris Convention if early promise is borne out. Many patent applications are successful despite the complexity and international divergence in the way they are examined. Some innovators choose to file numerous patent applications, and only pursue a proportion of them to grant. The selection process will reflect how the technology evolves in the early stages. The existence of patent applications, even if they are not ultimately taken forward, can still be a really effective means of warding off competition.

What about enforcement?

A major concern for those without deep pockets, like start-ups and academic institutions, is the cost of enforcement. Bringing patent infringement proceedings can be an expensive business, especially when more than one country is involved. Suing a larger opponent can be very daunting, and tactics to delay proceedings and make them more difficult can be rolled out by well-funded opponents. Short-form litigation like London’s Intellectual Property Enterprise Court offer a streamlined option in some situations. And insurance can be a tool to help manage risk and costs.

Dissatisfaction with the current system

Clearly, however, there’s still some dissatisfaction with the current system. Overall, 50% of the participants in our workshop felt that a new form of protection with a shorter life span is needed. Another 25% took the view that the whole patent system needs to be reviewed and brought into line with the 21^st century world.

While patents may be the best form of protection currently available for new methods and techniques, access to and control over the underlying data is also key. Some feel that the dominance in some areas by a small number of major players leaves others at a serious disadvantage. Ownership of data is a thorny subject, with variation between countries and a lot of uncertainty. Clear and comprehensive agreements between collaborators can make all the difference here, alongside diligent protection of confidential material, but big players are able to build on their head start with access to data.

UK IPO Consultation

The discussion was timely – on 7 September the UK Intellectual Property Office published a consultation on how the law can be adapted and updated to reduce uncertainty and provide a better fit for innovative AI technology, due to close on 30 November. We will be developing a submission in response to this consultation, bringing together opinions from our clients and contacts. Indeed, addressing the Chartered Institute of Patent Attorneys Congress this week, Tim Moss, Chief Executive of the UK IPO, specifically encouraged firms to collect comments from their contacts on how businesses would like to see the intellectual property system adapt to AI.  We would welcome your views.  

Next event - Big Data and IP

We will be running a similar online event on Big Data and IP in early November. If you would like to be invited to this, and future events, please get in touch.