The history of patents dates back to at least the Venetian Patent Statute of 1474 and arguably even further. During that time, the patent system has often struggled to come to terms with new technologies, as can be seen from the ongoing debate over computer implemented inventions. How to deal with artificial intelligence (AI) represents the latest challenge faced by the patent system and the UK Intellectual Property Office (UK-IPO) would like to know what you think about it by opening up a consultation on AI and IP.
With AI, a system of algorithms can be trained using data and the data effectively forms or adapts the algorithms. This approach has already been adopted, and has the potential to be genuinely transformative, over a wide range of applications.
New and innovative AI is potentially patentable, at least in the UK and Europe, as long as it is either specifically applied to a given technical application (applied AI) or is specifically adapted to the technical features of the system on which it is run, such as being adapted to better utilise different types of processors to perform calculations. However, inventions that never leave the sphere of theoretical mathematics, such as new generic AI algorithms, despite arguably requiring just as much if not more ingenuity, are currently unlikely to be patentable in the UK and Europe unless one of the above criteria are met, i.e. they are applied to a specific technical application or adapted to a specific technical system.
However, AI opens up a whole range of challenges for the patent system, which has its roots in the days of steam engines and other mechanical devices.
For example, what if the AI algorithm adapts itself based on the data it receives and generates a new innovative product feature. Who is the inventor? This is one of the questions for which the UK-IPO’s consultation is seeking opinions.
We have already seen patent applications that try to name AI algorithms as the inventor, and which have so far largely been rebuffed (for example, in the recent case Thaler v. Comptroller-General the Court of Appeal dismissed an appeal against the High Court's decision that an individual’s patent application which named an AI machine as the inventor was rightly rejected by the UK-IPO, because an inventor has to be a natural person under the Patents Act 1977). This is a controversial area as the inventor can be the first owner of the patent, so would having an AI algorithm as an inventor on a patent application also require allowing it to “own” the patent? That would be a step that many might find unpalatable.
The consultation also seeks views on copyright protection for computer-generated works without a human author. The UK is one of only a few countries to offer copyright protection (50 years from the date the work is made) for works generated by computer where there is no human creator – the author being defined as “the person by whom the arrangements necessary for the creation of the work are undertaken”. The consultation asks whether protection should be more or less than that offered at present, or indeed be available at all.
Another question that is frequently asked is if AI algorithms can be trained using training data that is publicly accessible, such as from the internet. Often the AI algorithm can “crawl” the internet and other digital sources looking for appropriate data, such as digital images, and training itself on that data. Often the internet is the only practical source of training data. However, that data might be protected by copyright and database rights, and it would be wise to keep that in mind when considering using third party data to train an AI algorithm.
The UK-IPO consultation also looks for input on this issue of text and data mining (TDM), asking if the existing UK copyright framework which has a specific copyright exception for TDM (introduced in 2014) needs updating to reflect the importance of these techniques in AI use and development. On one hand, broadening the exemption may facilitate AI development, particularly in areas where good training data is hard to come by. On the other hand, the collection and processing of such data into a suitable form can often require significant resource and give rise to significant competitive advantages in the resulting AI algorithms, and relaxing the current TDM exception might be seen as penalising those who invest that resource to create good training data over those that simply seek to exploit it.
Whatever the conclusions, it will be interesting to see the outcomes of the UK-IPO consultation and how they steer treatment of copyright and patenting in relation to AI. A settled and predictable IP system for AI can only be of benefits to creators, innovators and users of this transformational technology. The UK-IPO consultation on AI and IP runs to 7th January 2022, so something to ponder on over the festive period! Details can be found at: https://www.gov.uk/government/consultations/artificial-intelligence-and-ip-copyright-and-patents.