Discussions about artificial intelligence – and its relationship with intellectual property – often raise fascinating hypotheticals.
For much of the last two years or so those discussions have centred on whether an AI machine can be named as an inventor on a patent application. The answer so far, going by the letter of the law, is no.
But that doesn’t mean the possibility of AI-created inventions is a non-starter. On the contrary, the general sense is that at some point we are going to have to start finding ways to accommodate AI-driven creations.
At the Westminster Media Forum last Friday, April 8, a mix of in-house IP counsel, academics and practitioners considered how a future UK regime could account for AI developments in the context of copyright.
The discussion, held online, took place while the UKIPO considers responses to a consultation it ran at the end of last year on IP protection and its relationship with AI.
Related stories
Mining exemptions
The main talking point of the day was around how copyright law could adapt to allow for use of AI and what, if anything, needs to change in the law. UK copyright law is currently governed by the Copyright, Designs and Patents Act 1988.
Much of the conversation considered the need for an exemption in copyright law to allow for text and data mining (TDM). TDM broadly refers to the automated process of selecting and analysing large amounts of text or data to find patterns, discover relationships, or create new ideas.
The development and training of AI requires large swathes of data to be fed into machines and analysed, delegates noted. However, they pointed out that access to much of that data is often protected by copyright or database rights.
Unlike in patent law where attempts to list an AI tool as an inventor have so far failed, there is hope that AI can be considered the author of copyright-protected work.
The Canadian Intellectual Property Office and Indian Copyright Office have both granted a copyright registration recognising an AI painting app as a co-author. The AI machine, called RAGHAV Artificial Intelligence Painting App, used Vincent van Gogh’s painting ‘Starry Night’ and a photograph taken by Ankit Sahni – an IP lawyer and owner of the AI machine – as base datasets to create the painting.
One argument put forward at the conference that could help encourage AI inventiveness is a specific TDM exemption, which delegates said should be outlined under UK law.
In the EU, the Copyright Directive, which member states are in the process of transposing into their national laws, allows use of TDM without the permission of the relevant copyright owners (though rights owners are given the possibility of defending their assets).
As it stands, the UK has no such provision. In response to the government’s first consultation on AI at the end of 2020, industry association the IP Federation suggested that a broad TDM exception should be adopted.
The government has since launched a second consultation, which closed in January this year, seeking additional views on how to make it easier to use copyright-protected material in AI development and training as well as IP protection for inventions made by AI.
However, conference attendees were split on the idea of a TDM exemption.
Christie mystery
Arthur Heard, senior legal advisor at publishing company HarperCollins in London, posed a hypothetical scenario of an AI machine that ingested only works by Agatha Christie and created a new novel.
He posed the question: “Under current legislation the owner would be the person who programmed the AI, but if copyright is supposed to be protecting the expression of idea, is there an argument that Agatha Christie [or her estate] should own some of the copyright?”
He noted that the source material may have been copied, even if exact phrases were not used, and that in the absence of clear guidance or legislation, it may come down to case-specific examples in future.
According to Heard, licensing is the best possible model to avoid these questions.
“Publishers are in the business of making money and are open to licensing. Without money to reinvest in authors and remunerate them for their work, you run into problems later on,” he said.
But Lorna Campbell, trustee at Wikimedia – the owner of online encyclopaedia Wikipedia – and an academic at the University of Edinburgh, said she is in favour of a “simple and straightforward universal education exemption” that would allow TDM for educational purposes.
She added that the internet, which transcends national borders, risks becoming more fragmented unless copyright rules and norms converge.
“In considering reform we need to take into account everyone’s needs, not just those of multinationals but also teachers and learners.”
Campbell added that legislation should be reshaped to fit modern education with all publicly funded educational resources free and open to the public. Open licences, as well as a clearly defined fair use policy similar to that deployed in the US, would help, she suggested.
One of the most notable judgments centring on TDM and copyright was the Authors Guild v Google case in which both the US District Court for the Southern District of New York (2013) and the US Court of Appeals for the Second Circuit (2015) found that Google Books’ harvesting of textbooks was fair use and transformative.
Scott Roberts, head of patents at BT and vice president of the IP Federation, said introducing a TDM exemption must be a priority for the government “at least on the same level as in the EU”.
Returning to the Agatha Christie example, he suggested that influence has long been a factor in the creation of new works.
“Christie herself acknowledged influences, including [Sir Arthur] Conan Doyle. There are lots of similarities between Doyle and Christie’s work. I’m not sure if AI should be hamstrung by not being able to digest publicly available materials as a way to create commercial outputs.”
He added that the training of AI – including extensive digestion of data – is essential for it to become efficacious, which enhances the argument in favour of a TDM exemption.
Roberts highlighted two problems that need to be solved to ensure UK law is able to fully account for the development of AI: adequate access to good data through exemptions, and an assurance that innovations arising from AI can have IP protection.
“This has to be done in an internationally harmonised way and it’s not going to be easy,” he argued.
Earlier in the conference delegates heard from Tim Moss, chief executive of the UKIPO.
Moss noted that that the office’s three-year strategy, due to be published in the next few months, will broadly set out three goals: delivering core services, creating a world leading IP environment, and making the UKIPO a “brilliant place to work”.
The day’s discussions will have given the office plenty of food for thought as it considers how to make those goals a reality.
