'AI and Performers’ Rights in Historical Perspective' (CREATe Working Paper 2023/9) by Elena Cooper comments
This article uses legal history as a vantage point for reflecting on the current moment in the debate about AI and performers’ rights. Current debates often refer to ‘creators’ and/or ‘copyright’ as generic categories denoting both performers and authors. Legal history, I argue, sharpens the critical lens on current debate by drawing our attention to what today remains different about the legal rules protecting performers. That difference, at present, leaves performers less well placed to deal with the challenge of AI than authors and also goes to the heart of Equity’s current reform proposals. That difference should now be debated.
Last year, I published an Opinion in the E.I.P.R. - Copyright History as a Critical Lens – a follow- up to Interrogating Copyright History (an E.I.P.R. Opinion that I co-authored with Ronan Deazley in 2016). I argued that the study of law in past times can be a powerful critical lens on how we see the legal present. Whether the past reveals a story of continuity or change, there is, I argued: value in looking backwards, before we look forwards: an historical perspective helps us to recover the contingency of the present, to imagine things differently and to look to the future with a more critical eye.
My comments should be read in the light of a wider ‘historical turn’ in critical thinking about intellectual property law in the last two decades, following the publication in 1999 of The Making of Modern Intellectual Property Law by Brad Sherman and Lionel Bently. Sherman and Bently’s historical work showed that the categories of intellectual property that we know today, are not timeless, natural or inevitable; ‘theory... played at best an ex post facto role’ in later legitimating the legal categories that emerged. In so doing, Sherman and Bently opened the way for legal historians to probe critical questions about the legal present and future of intellectual property law.
In this article, I provide an example of looking backwards before we look forwards. I demonstrate the critical value of an historical long-lens on a discrete strand of current legal debates raised by cutting-edge technology today: a facet of the impact of Artificial Intelligence technology (or ‘AI’) on performers, particularly actors, and the search today for an appropriate legal response. Equity, the UK actors’ trade union launched a campaign last year, Stop AI Stealing the Show, seeking the legislative reform of statutory performers’ rights, specifically the increase in the scope of legal protection. However, the UK Government has, so far, resisted reform. The UK Government’s position on performers’ rights is indicated in the closing paragraphs of its response to the UK Intellectual Property Office’s consultation AI and IP: Copyright and Patents. Referring to proposals for ‘an expansion of the scope of performers’ rights in the Copyright, Designs and Patents Act 1988’, the UK Government comments as follows: at this stage, the impacts of AI technologies on performers remain unclear. It is also unclear whether and how existing law (both in the IP framework and beyond it) is insufficient to address any issues. If intervention is necessary, the IP system may not be the best vehicle for this. We will keep these issues under review from an IP perspective.
How might an historical perspective enable us critically to reflect on the present moment in the debate of the future of performers’ rights?
Technology in the 21st century: Aspects of the challenge of AI today
Before I look to the past, I start with today. AI is technology that uses machine-based learning to perform functions that were previously the province of usually slower, more costly and/or more labour-intensive processes undertaken by human beings. AI has a wide field of application from medical diagnosis, robotics, to the management of insurance-risk. However, one set of questions for intellectual property law today relates to AI’s impact on authors and performers working in a variety of sectors.
The creative potential of AI technology has been embraced by some visual artists and AI has been hailed elsewhere as democratising creativity, by providing everyone with the tools to create cultural works. AI technology also offers new possibilities for enhancing human performances, for instance, in the creation of video games, in turn opening opportunities for actors. Yet, there are also reports that AI is, in certain contexts, increasingly replacing human authors and performers, and putting them out of work. A good example of this is the audio performance sector, where AI generated voices can now be used to undertake audio work (e.g. audio books) or provide voice-overs at negligible cost, in circumstances where a professional actor would previously have been employed. Redundancy for actors caused by the introduction of AI – the replacement of humans by machines – is reported to be now increasingly commonplace.
While AI can replace human authors and performers, it also frequently utilises their pre-existing work. AI learns by tracking patterns in an existing body of material: a ‘data-set’. In the case of visual images produced by AI, the data-set may comprise large quantities of copyright- protected visual images scraped from the internet without copyright clearance. For AI produced voices, the data-set may be a collection of recordings of human voices, which may have been recorded by actors for another unrelated purpose (e.g. a casting or audition) and included in the data-set without consent. Alternatively, the data-set may comprise recordings that were licensed to a third party for broad purposes, e.g. ‘for research’, yet particularly where the licence pre-dates AI technology, AI uses were not specifically contemplated by the parties at the time the contract was concluded.
In addition to the AI learning process, the material generated by AI – for instance images generated in response to a text prompt – may involve ‘copying’ through a new means: computer synthetisation. In relation to performance, prior to AI technology, the circumstances in which a performance could be copied, without direct taking from a recording itself, were more limited and confined to human imitation such as a sound-a-like imitating an actor’s voice, as in the passing off case of Sim v Heinz (discussed further below). By contrast, AI technology today opens a future in which a performance, or aspects of a performance, can be recreated through technology, without direct copying from the recording. Equity, adopting the arguments of Mathilde Pavis in Artificial Intelligence and Performers’ Rights, refers to these new modes of copying performances, via ‘digital sound and look-alike’, as ‘performance synthetisation’.
Whether or not performers are sufficiently protected is, of course, tied to the distinct circumstances raised by the new and unprecedented technologies of today: the challenge for legislators and the courts is to strike the right balance of interests in a specific technological context and then (as in the case of the Economics of Music Streaming Enquiry in recent times) to continue to track how well that ‘balance’ operates in practice. How, then, can the past help us to reflect on debates about AI and performers today?
