Jake Elwes’ Terms & Conditions Opera: a Legalese Libretto (2024–2025) turns the fine print of corporate AI platforms into an absurdist operatic score. Elwes constructed a process whereby OpenAI (famous for ChatGPT) and Suno (a music creation AI) could be fed their own usage policies and terms of service. In this self-referential loop the systems absurdly reinterpret the legal frameworks that govern them, producing a kaleidoscopic blend of musical genres. For Elwes, the true content of corporate AI lies not in its generated output – often imitative and derivative – but in the policies, copyrights, and terms that shape its use and demand real human labour and political engagement. By “feeding an AI on itself” the work renders opaque legal jargon strangely legible and even laughable, highlighting the ethical, legal, and creative tensions at the heart of our relationship with machine-generated culture. Terms & Conditions Opera connects directly to the November 2024 manifesto and collective action Elwes played a central role in, under the title ‘Art in the Cage of Digital Reproduction’, in which artists withdrew their labour from OpenAI’s then unreleased video generation model, Sora.

A music AI being fed on its own legalese. In a self-referential loop, OpenAI and Suno's Terms and Conditions are reinterpreted into a kaleidoscopic opera of scattered genres.

The true content of corporate AI systems lies not in their generated output (which are merely imitative, superficial and derivative forms), but in their legal frameworks: the policies, copyrights, intellectual property, and terms of usage that demand real human labor and political engagement.

Prompting corporate AI models can feel very limited and troubling - almost as if you’re giving part of your soul to help improve their models. Not wanting to give their model anything of myself, I instead fed it on itself, having it absurdly interpret its own governing laws.

AI is forcing us to radically reinvent intellectual and creative property laws (both OpenAI and Suno are currently being sued for copyright infringements). The revisions being shoehorned into this newly evolving landscape are buried in boring documents none of us read, but when set to mimetic music their jargon becomes both more readable and laughable. Decadently feeding an AI on itself (a de-generative AI) resulting in an absurdist AI opera.

Jake Elwes (b. 1993) is a conceptual artist, hacker, radical faerie and researcher living in London. They have been making critically engaged art exploring the aesthetics and ethics of machine learning systems since the very first generative AI models in 2016. Across projects that encompass moving-image installation, sound and performance, Elwes’ work finds unusual ways of demystifying, mapping and subverting technology. Their work searches for poetry and narrative in the successes and failures of digital systems. Works include deepfake drag in The Zizi Project, glitching oppressive algorithms in Machine Learning Porn and reframing AI generated marsh birds back into nature in CUSP. Their work also calls for us to challenge who builds these systems and for what purpose, and whether artists can reclaim these technologies to build their own digital utopias. Elwes’ work has been exhibited internationally, including at the Victoria and Albert Museum, London; UK; Pinakothekder Moderne, Munich, Germany; ZKM, Karlsruhe, Germany; Today Art Museum, Beijing, China; Fundacion Telefonica Museum, Madrid, Spain; Museum für Naturkunde, Berlin, Germany; and Nature Morte, Delhi, India.