When the Royal Shakespeare Company's production My Mum's a Twat used live facial-capture technology to translate actress Kathryn Hunter's expressions onto a digital avatar in real time, critics split sharply. Some called it visionary. Others asked: whose performance was the audience actually watching? Hunter's body generated every micro-expression — but the face the audience saw belonged to a generated image. The debate cut to the heart of what live performance actually is.
Performance theorist Philip Auslander argued in his 1999 book Liveness that "live" performance is not ontologically prior to mediatised performance — both are culturally constructed categories. But practitioners from Jerzy Grotowski to Peter Brook maintained something irreducible exists when a trained human body inhabits a shared space with an audience: risk, breath, the possibility of failure.
Neuroscientist Vittorio Gallese's research on mirror neurons — published in journals including Trends in Cognitive Sciences — showed that watching another human body in skilled motion activates motor resonance in the observer's own body. We literally feel, at a neural level, what the performer does. This embodied empathy is fundamental to why live performance generates emotional responses that filmed performances often cannot fully replicate — and it raises the question of whether audiences can achieve the same resonance watching an AI-animated body.
In 2019, researchers at the Max Planck Institute for Empirical Aesthetics published findings showing that audiences reported significantly lower emotional response to identical theatrical content when they were told the performer had been replaced by a motion-capture avatar, even when they could not visually distinguish the two. The mere knowledge of human absence reduced engagement.
Max Planck Institute, 2019: Audiences told a performer was AI-animated rated identical movements as less emotionally resonant — even when they admitted they could not detect a visual difference. The human signal matters beyond its surface appearance.
Jerzy Grotowski's concept of Poor Theatre — stripping performance to the bare actor-audience relationship — defines the performer's body as the singular irreplaceable instrument. Grotowski wrote in Towards a Poor Theatre (1968): "The actor who undertakes an act of self-penetration… offers himself." This offering — the risk of genuine exposure — is what AI systems, trained on past human data, categorically cannot do. They hold nothing at risk.
Yet in 2023, the Guthrie Theater in Minneapolis integrated GPT-4 into its dramaturgy process, using the model to generate alternative scene interpretations and character psychology notes for its production of Hamlet. The human actors and director then filtered those notes through their own judgment and experience. The AI contributed analysis; the humans contributed presence. This division of labour is being adopted increasingly across regional theatre.
AI can generate text, animate faces, compose movement — but cannot place itself at genuine risk. Performance, in its deepest sense, requires something to be at stake for the performer. This is the human question AI has not answered.
You've learned about embodied empathy, mirror neurons, and Grotowski's Poor Theatre. Now push deeper. Use the AI assistant to explore: what does human presence actually consist of, and which elements — if any — could ever be replicated by AI systems?
When choreographer Wayne McGregor premiered Living Archive at Sadler's Wells in 2019 — having used machine learning trained on forty years of his own movement vocabulary to generate new choreographic material — the programme credits listed McGregor as choreographer. The AI system was described as a "creative tool." Dancers said the system's suggestions felt alien yet generative: it extended McGregor's idiom in directions he would not have reached alone. Yet the final work was filtered entirely through his taste and the dancers' bodies.
In February 2023, the US Copyright Office issued guidance clarifying that works produced entirely by AI without human creative selection cannot be copyrighted. In Thaler v. Vidal (Federal Circuit, 2022), the court held that AI systems cannot be listed as inventors on patents. These rulings establish a consistent principle: US intellectual property law recognises only human authors.
However, the cases become murky when human creative selection is involved. The Copyright Office's guidance noted that "the extent to which there is human authorship" in an AI-assisted work determines copyrightability — but gave no precise threshold. This ambiguity is already appearing in performing arts contracts. The Writers Guild of America's 2023 strike agreement with major studios explicitly restricted AI's role in scriptwriting and required disclosure of AI use, establishing precedent that human authorship must be protected by contract even when law remains unclear.
In the UK, the Copyright, Designs and Patents Act 1988 contains a provision — Section 9(3) — that does recognise computer-generated works, attributing authorship to "the person by whom the arrangements necessary for the creation of the work are undertaken." This makes the UK an outlier internationally and has drawn increasing scrutiny as AI capabilities have expanded.
WGA 2023 strike agreement: AI cannot write or rewrite literary material; producers must disclose if they provide AI-generated material to writers; AI-generated material cannot be considered a writer's "professional writing sample." This is the first major US labour agreement to directly regulate AI in creative work.
Legal authorship and artistic credit are not the same thing. When Holly Herndon and Mat Dryhurst released the album PROTO (2019) — which trained an AI they named "Spawn" on ensemble vocal recordings and used its outputs in the final work — they credited Spawn as a collaborator in press materials, even though the AI held no legal copyright. This was a deliberate ethical and conceptual statement about how we should understand collaborative creative processes with AI.
The opposite approach appeared in 2023 when the estate of George Carlin sued the creators of a podcast that used AI to generate a "new" George Carlin comedy special, arguing the system replicated his voice, style, and persona without consent. The creators settled. The case raised questions not just about copyright but about dignity rights — the right of a performer's estate to control posthumous AI-generated performances.
McGregor has described the AI as "a mirror trained to reflect my movement language back at me in unexpected configurations." The question of authorship, he suggests, is less about who or what generated the raw material and more about who exercised taste, judgment, and took responsibility for the finished work. That person was him.
The authorship question is legally unresolved and artistically contested. Use this lab to test the limits: push the AI assistant on edge cases, thought experiments, and real scenarios where the boundaries of creative credit become genuinely difficult to draw.
During the 2023 SAG-AFTRA strike, one of the central disputes was the use of AI to scan actors' likenesses for perpetual and unlimited use with a single day's pay. Studios had begun offering background performers a flat fee to capture their digital likeness — effectively replacing that performer with a replicable AI copy for any future production, indefinitely, for a one-time minimum wage payment. SAG-AFTRA called this "the soul of the strike." The final agreement included protections requiring informed consent and residual payments for AI-generated likenesses.
A 2023 report by the Actors' Equity Association surveyed 1,400 working stage actors. 67% reported concern that AI voice and motion-capture technology would reduce the number of available performance contracts within five years. 41% said they had already lost or were competing with AI-generated content for specific gigs — particularly in voiceover, audio drama, and motion capture for video games.
The video game industry is particularly instructive. The National Institute for Labor Relations Research documented that between 2021 and 2023, major studios including EA and Ubisoft significantly reduced voiceover casting calls, replacing some work with AI voice synthesis. In response, SAG-AFTRA launched a dedicated "No AI" campaign in 2024 targeting game studios, and negotiated the Interactive Media Agreement covering minimum AI use standards.
The live theatre ecosystem faces different but related pressures. AI-generated music, originally composed for a production's needs at a fraction of traditional commissioning costs, has begun appearing in small-budget regional productions. The American Federation of Musicians reported in 2023 that live pit orchestras for touring Broadway productions had declined 23% over the previous decade — a trend that began with pre-recorded "virtual orchestras" but is now accelerating with AI composition tools like those offered by AIVA and Soundraw.
Informed consent required for digital likeness capture. Performers must receive residual payments if AI-generated likenesses are used. Studios cannot alter a performer's likeness to make them say or do things they did not agree to. AI-generated "digital doubles" require the same compensation as the performer themselves would receive.
Not all analysis is pessimistic. The National Endowment for the Arts' 2023 report AI and the Creative Economy noted that AI tools have significantly lowered the production cost barriers for independent artists and small companies. Virtual production tools powered by AI have enabled productions at regional companies like the Oregon Shakespeare Festival to achieve visual effects previously affordable only to major studios — potentially expanding the range and ambition of work that can be produced with smaller budgets.
Theatre practitioners including director Rachel Chavkin have publicly argued that AI's greatest potential contribution is eliminating the administrative labour that consumes artists' creative time — scheduling, budget modelling, grant writing — rather than replacing performance itself. In 2023, Chavkin noted using AI tools to draft initial grant applications, freeing her team to focus on artistic development. This is a pattern repeated across many companies: AI used for infrastructure, humans retained for creation.
AI reduces the marginal cost of producing certain types of performance content toward zero. In a market economy, this means whoever owns the AI tools captures the value that previously went to labour. Without contractual protections, the efficiency gains of AI flow to capital, not to the artists whose work trained the systems.
The labour questions are where the human stakes become most concrete. Use this lab to explore: what protections should exist, who benefits from AI efficiency gains, and what the performing arts community should demand in this transition.
At the 2023 Edinburgh Festival Fringe, the company Uninvited Guests premiered How to Occupy an Abandoned Space — a performance that used live AI text generation projected in real time as one "voice" among three. The audience saw the AI generating responses to the performers' prompts on screen; the human performers then responded to the AI's output, visibly deciding what to take and what to discard. The process of human discernment was itself made visible as the artistic act. Reviews described it as "the most honest use of AI in theatre yet seen at the Fringe."
The Uninvited Guests model points toward a principle that has emerged across several practitioner conversations and manifestos in 2023–2024: transparency about AI use is itself an artistic and ethical commitment. When audiences know what is human and what is generated, they can make informed decisions about where to direct their attention and empathy.
In 2023, the Theatre Communications Group (TCG) — the primary service organization for US nonprofit theatre — published a statement on AI in theatre that included a call for disclosure norms: productions should indicate in their programme notes whether and how AI was used in creation or production. No major institution has made this mandatory, but several leading companies including the Berkeley Repertory Theatre and Arena Stage began voluntarily disclosing AI involvement in 2024.
Similar principles are emerging internationally. The Australian Theatre Forum published AI Ethics for Theatre Makers in early 2024, identifying four core commitments: transparency, consent, fair compensation, and human artistic authority. These echo the SAG-AFTRA provisions but extend them from labour rights to artistic principles.
The Theatre Communications Group called for: (1) industry-wide disclosure norms for AI use in production; (2) protection of human authorship and labour; (3) use of AI as a tool that amplifies rather than replaces human creativity; (4) ongoing community dialogue about values. No enforcement mechanism — but normative framing matters.
Drawing on the cases in this module, a working framework for performing arts practitioners integrating AI can be organised around four questions:
1. Who holds artistic authority? The human practitioner must retain final creative judgment. AI generates options; humans exercise taste, values, and responsibility. This is not merely an ethical preference but the basis for legal authorship.
2. Who consents and who benefits? Any performer whose likeness, voice, or movement is captured for AI training or replication must give informed consent and must benefit from subsequent use. This principle is now contractually established in the SAG-AFTRA agreement and is the minimum floor.
3. What is disclosed? Audiences, collaborators, and employers have an interest in knowing how AI was used. Transparency about AI contribution enables informed response and prevents deception.
4. What is preserved? Certain elements of performance practice — live presence, physical training, the actor-audience contract — should not be surrendered to efficiency. Practitioners must actively decide what is irreplaceable and protect it.
Theatre has absorbed disruptive technologies before: electric lighting in the 1880s transformed staging; recorded sound in the 1920s threatened live music; cinema in the early twentieth century was widely predicted to end theatre entirely. Each time, practitioners found that the disruption clarified what live performance actually was and why it mattered. Electric light enabled expressionism. Cinema's narrative power pushed theatre toward forms cinema could not replicate.
The AI disruption follows this historical pattern — but moves faster and strikes more deeply into the cognitive and creative functions that practitioners previously considered distinctly human. The performing arts community's response will shape not just theatrical practice but broader cultural understanding of what human creativity means and why it matters.
This module has not resolved the human question — it has mapped it. The question of what makes performance irreducibly human will be answered not by philosophers or technologists, but by practitioners making specific choices in specific rooms with specific audiences. That ongoing practice of discernment is itself the answer.
You've worked through four lessons covering presence, authorship, labour, and practice frameworks. Now synthesise. Use this lab to construct your own position on how AI should be integrated into performing arts — and have the AI assistant challenge it.