By Ivan Yuen Man Chun, Research Director of the King’s Commercial Awareness Society
The UK government claims to be “putting the UK on course to be the best place in the world to build, test and use AI technology”. This claim is not without its merits, as the UK is already the third-largest AI market in the world after the US and China. The UK’s AI market is currently valued at over USD $21 billion, with even more significant growth expected to occur in the coming years.
However, how the government approaches the contentious and pressing issue of applying copyright law on artificial intelligence can be decisive in regard to the UK’s ability to achieve its AI market ambitions. Therefore, the government accepted a recommendation by Sir Patrick Vallance, the Chief Scientific Advisor of the UK, to clarify the relationship between intellectual property and generative AI, by working with intellectual property rights holders on a code of practice on copyright and AI.
The code of practice must provide clarity on two critical questions: whether AI learning (text and data mining) (TDM) on the complete works of a human creator without their express permission should be considered a copyright infringement, and whether works created with AI assistance should be copyrightable, and if so who the copyright should belong to. The government must approach these matters in a manner that balances AI’s need for greater regulatory freedom to ensure market growth and content creators’ need for adequate legal protection of their ideas and innovations to maintain their incentive to create and invest in R&D.
Should AI Learning on Complete Works be Considered Copyright Infringements?
In June 2022, the UK Intellectual Property Office (IPO) announced a proposal to allow TDM for all purposes whatsoever. This exception to British intellectual property law would allow commercial AI tools to be trained on all copyright-protected works without the need for a licence, rightsholder permission, or compensation.
This instigated widespread fears that the proposal would likely devastate creators’ incentive to create and share original work, allow large corporations wielding powerful AI tools to rapidly learn from, improve upon and replace contributors of creativity to society, such as artists, authors and designers, and allow said large corporations to utilise those ideas that did not belong to them to eventually extract a commercial profit, leaving the original human creators with no credit for providing the foundation on which the AIs were trained.
Unsurprisingly, the significant objections from rightsholders caused the government to abandon this proposal in February 2023. However, a blanket ban on AI from training on all copyright-protected works contains another set of ramifications. Restricting AI learning material to non-copyrighted work severely impedes the speed and breadth of its learning. This majorly hinders AI’s functionality and utility, risking the stagnation or even decline of the UK’s rapidly growing AI market, rendering the UK’s AI ambitions virtually unachievable.
What is the appropriate compromise? The EU does so through Article 4(3) of the EU Digital Single Markets Directive, which allows TDM on copyrighted works but provides creators with the option to expressly reserve their rights, and essentially opt out of TDM. An alternative may be a financial compensation agreement between AI companies and rightsholders to rightfully credit them for their work’s contribution to AI learning, although the financial cost of all compensation packages for all copyrighted work may prove unrealistic to afford.
Nevertheless, a middle-ground solution by the code of practice is most likely to balance AI market growth and rightsholders’ intellectual property rights.
Should AI-assisted work be copyrightable? If so, who should own the copyright?
In current British law, computer-generated work can only be protected by the Copyright, Designs, and Patents Act 1988 (“CDPA”) if it is “original”, but the test for originality is currently uncertain, rendering the current law on the matter unknown.
Nonetheless, allowing all computer-generated works to be copyrighted by the human author that prompted the AI regardless of the minimal intellectual effort made arguably defies the very meaning of intellectual property. It is debatable whether a ‘creator’ inserting a single sentence prompt input into an AI tool deserves credit and copyright ownership over the ‘intellectual property’ the AI generates.
More deserving of the copyright, however, maybe the AI software developer. The less effort expended by the author, the greater the proportion of credit that should belong to the software developers of the AI. Granting copyrights of largely computer-generated works to AI software developers would create a great incentive for AI developers, potentially accelerating market growth. This would be more beneficial than rendering AI-generated work non-copyrightable.
However, most commercial AI usages do not rely wholly on AI generation. AI generation is typically utilised to assist and inspire new original human creations. This would typically pass the old English law test for originality, “skill, judgment, and labour”, providing a stronger justification to grant copyright to the human authors who merely utilised AI as secondary assistance to their original work.
This is analogous to Hyperion Records v Sawkins [2005] EWCA Civ 565, where a composer and musicologist created new versions of a public-domain work by including corrections and additions. The Court of Appeal found that the composer’s revisions made it an “original” work. Such granting of copyright would encourage creators of various works to frequently utilise AI assistance without the fear of losing copyrightability, fostering a cooperative and co-dependent relationship between AI and human content creators rather than a competitive and binary one, simultaneously satisfying content creators’ interests and stimulating the AI market in the UK.
Looking Forward
China and the US have already clarified their positions on AI copyright regulation. This gives the UK further urgency to clarify the code of practice and stabilise the legal and commercial uncertainty surrounding copyrights involving AI. It is evident that a compromise between AI’s need for greater copyright regulation freedom and creators’ need for adequate legal protection of their innovations is the UK’s best chance at achieving their AI market ambition while equitably protecting creators’ intellectual property. Content creators should be given incentives to work with AI, such as strong copyright to their works, to optimally achieve the harmonious coexistence of human creativity and artificial intelligence.
