Generative AI continues to make advances. And whilst its capabilities can be overhyped, there is undoubtably a growing perception that AI will soon be capable of effectively and infinitely ‘replacing’ the human performers on whose performances it is trained. Whether AI will ever fully achieve this goal in a cost-effective manner, or if the market will embrace this development, is yet to be determined – but the immediate threat to the performing arts is clear. Back in 2023, the risks of AI-displacement formed one of the core issues of the SAG-AFTRA strikes, the largest withdrawal of creative labour in Hollywood history. More recently in 2024, the high-profile dispute between Scarlett Johansson and OpenAI’s chatbot ‘soundalike’, which trained an AI on an unknown actress’ voice who sounded tellingly like Johansson after the star refused to allow them to use her voice, shows that not even the highly privileged Hollywood elite are exempt from automation.
The potential harms of ‘replacing’ the (creative) human can be myriad. Economic harms caused by labour displacement threaten mass redundancies in an already tenuous sector. Social and cultural harms can also arise where existing performances can be repurposed, e.g. to create deepfakes of a single person or to create a stereotyped version of a particular community. Of course, all of these potential harms will have disproportionate effects on communities which are already vulnerable or marginalised, e.g., where AI can be leveraged for revenge porn or whitewashing.
In short, despite its ostensible technical neutrality, powered by ‘value neutral’ tech based on mathematical and computational theory, AI nonetheless raises highly divisive and political questions about representation, diversity, and sustainability. Among these is the ultimate question of what value we give to performers, and correspondingly, whether the legal protections they are currently afforded offer them appropriate safeguarding mechanisms. This is an issue we explored recently in a co-authored an article for the EIPR’s AI Special Issue (also available as a CREATe Working Paper): Evidencing the Value of Human Performance. In this article, we brought together multiple perspectives on performers’ rights in the AI-era, combining legal historical research (Elena Cooper), empirical studies (Amy Thomas), and real-world experiences (Laurence Bouvard).
In approaching this research, our guiding principle as to what a well-configured performers’ rights system should look like, with or without AI, is one that promotes cultural diversity, both in the sense of how the configuration of performers’ rights can (i) enable diversity of participation in cultural markets and (ii) enable diversity of outputs in cultural products.
On the first point, our review of the available empirical evidence (based on the catalogued studies on the Copyright Evidence database and preliminary survey findings from a survey of UK-based audiovisual performers) suggests that ‘typical’ creative labour, performances included, trend towards being undervalued. Creative labour markets are characterised by unstable pay and working conditions, as well as ‘winner-take-all’ dynamics where only a relative few ‘superstars’ command a disproportionately large share of overall net earnings. This has an impact on diversity of all kinds, as only those performers who can square the paradox of having an alternative, reliable source of income, that simultaneously allows them to be continually available for whenever opportunities for performing work do arise, can afford to build a career. Whilst this issue is dealt with differently by different performers, the reality is that it is most easily solved either by having support from a wealthy family or by leveraging higher education qualifications in other, more lucrative fields. Those who can not benefit from either of these scenarios to supplement their income with alternative sources are often crowded out or prevented from participating in the first instance.
However, this reality is not necessarily an inevitable one (hence our article’s tag of ‘re-thinking’ performers’ rights). Indeed, one of the biggest factors dictating the structure of creative markets are the contractual terms offered by key gatekeepers – namely, recording companies, production companies etc. These contracts often consist of boilerplate terms, and are offered on a non-negotiable ‘take it or leave it’ basis. Agreeing to these terms can impact the extent of rights given to the gatekeeper, and resultantly can curtail long-term earning potential for performers. This can be especially so where ‘buy-out’ contracts (the transfer of all rights to a performance) are increasingly offered as standard by (in particular) streaming services. In these unbalanced relationships of power, policy makers in the UK (and elsewhere, particularly the US, given the ongoing SAG AFTRA video game voice-actor strikes) have explored ways to ‘even the playing field’ through protective mechanisms for creators when transacting with their rights, including mandatory fair remuneration, transparency obligations, and reversion rights.
This is of course directly connected with the second point: if we can secure inputs from a diverse range of performers, who are meaningfully enabled and sustained economically, we can also promote outputs that showcase a more diverse and attractive cultural landscape in performances that are more reflective of the wide range of human experience and stories. This is in some ways a key differentiator between performers and other types of creative labour where a ‘work’ is made: for performers, they are the product – diversity of input (of the performer) and output (of the performance) are inherently interlinked in a far more intimate way than other artistic endeavours such as writing, where the work is removed from its creator. If we seek improved representation of certain demographic groups on screen, the performer in that demographic group must be present in the performance.
However, the nature of AI is to reproduce the ‘norm’ based on a prompt, determined by a detectable pattern of what is ‘typical’ in the training data. In other words, AI has the capacity to crowd out diversity by excluding ‘outliers’, and the kinds of diverse, spontaneous instances of creativity that we should value. This raises an important contention of our article: AI and performers’ interests need not be oppositional. It is possible that a diverse, sustainable market of human performers can in fact improve and diversify AI performances by providing rich training data. But the terms on which that data is generated must be fair – especially against the backdrop of the (often inequitable) market dynamics illustrated in our first point.
The policy context of the UK (where the Beijing Treaty is yet to be ratified) makes for a particularly imperative base for research. In the UK, performers’ economic rights can be assigned to producers without any mechanisms for reversion, and their moral rights can be waived meaning they cannot be subsequently enforced. This means there is little ‘trade off’ between pay and external recognition for performers, recalling the finality of Don Draper’s exclamation: ‘that’s what the money is for!’ – a quote also utilised during the SAG-AFTRA strikes. This means that contract is again a likely area for policy intervention, especially if, as it is at least in the UK, performers’ can ‘sign away’ rights in their performances for the purposes of developing generative AI, without recourse.
This brings us to the conclusion that, whilst AI is certainly a disruptive technology in the performers’ market, it is perhaps better understood as an accelerant to many pre-existing issues surrounding power dynamics, sustainability, and diversity. AI intensifies these discussions because, by threatening to displace the human, we are forced to consider the value of that which will be lost. And in doing so, we must also confront the reality that the value of performance has been systematically, historically, and empirically undervalued – long before AI.
________________________
To make sure you do not miss out on regular updates from the Kluwer Copyright Blog, please subscribe here.
