At the San Francisco Museum of Modern Art on 15 November, artists Holly Herndon and Mat Dryhurst joined curator Hans Ulrich Obrist for a discussion about artificial intelligence, cultural systems and what they call “protocol art.” The phrase shifts attention away from the novelty of a tool and toward the rules that determine who can participate, what data can be used and how creative value is shared.
That is a useful turn for artists. Much public debate about AI remains fixed on outputs: whether an image looks convincing, whether a song sounds human or whether a model can imitate a recognizable style. Protocol art asks a prior question: who designed the conditions under which those outputs become possible?
When infrastructure becomes the medium
SFMOMA described protocol art as work in which usually invisible systems—algorithms, applications, permissions and technical rules—become artistic material. The artwork is less a fixed object than a process built from instructions and open to collaboration with people and machines.
Herndon and Dryhurst have pursued this idea through projects concerned with identity, voice, data provenance and consent. Their work with Spawning, for example, seeks to help creators inspect whether material appears in public AI training datasets and to express preferences about training use. These mechanisms do not solve every legal or technical problem, but they make control visible and actionable.
What this means for artists
Artists using AI can document more than prompts. They can record the model and version, source materials, permissions, collaborators, selection process and later human edits. That record helps distinguish deliberate authorship from automated generation and gives curators, collectors and audiences a clearer account of how the work was made.
Artists can also treat consent as a design choice rather than an afterthought. If a project uses another person’s voice, likeness, archive or private data, permission should be defined before production. If audiences contribute material, the terms of participation should be understandable at the moment of contribution.
The most consequential insight is that artists do not have to remain downstream users of systems created elsewhere. They can build, contest and redesign the rules themselves. That does not make every protocol an artwork, nor does it remove the need for law, contracts and institutional accountability. It does expand the field of creative agency.
In this view, the central AI question is not simply what a machine can make. It is what kinds of relationships artists choose to encode around making—and whether those relationships preserve consent, credit and meaningful human decision-making.
ATIM approaches artificial intelligence with curiosity, independence, and balance. We explore its extraordinary potential to empower artists while examining the questions it raises around authorship, copyright, consent, creative identity, transparency, and human expression.
We are neither advocates for AI nor opponents of it. Our focus is how the technology is used—and what that means for artists.
Above all, our commitment is to artists. We look at every development through the lens of what can support, protect, and empower the creative community—identifying opportunities, examining potential risks, and providing artists with the information they need to make informed decisions about their work, their rights, and their future.