Artificial intelligence is moving beyond the familiar prompt-and-response exchange. The next creative question is not only what a system can generate, but what it may be allowed to decide and do.
That shift is the focus of Agentic Aesthetics, a Rhizome program at the New Museum in New York on October 28, 2026. The event brings together artists Ian Cheng, Lynn Hershman Leeson, Gene Kogan and Franco Mattes, with Rhizome artistic director Michael Connor moderating. A free livestream extends the discussion beyond the museum.
Rhizome describes agentic AI as systems that can act on their surroundings and on user input with a degree of autonomy. The distinction matters. A generator produces an output after a request. An agent may pursue a goal through several steps, select tools, call services, revise a plan or initiate an action. In a studio, those abilities could support research, versioning, production coordination and exhibition maintenance. They can also make responsibility harder to see.
From making an image to delegating a decision
Artists have long built rule-based, generative and autonomous systems. Cheng’s live simulations, for example, place software agents inside changing worlds rather than treating the computer as a passive instrument. Hershman Leeson’s decades of work with chatbots, avatars and networked identities similarly predates the current product cycle. Their presence gives the conversation a longer artistic history: agency in art is not new, but the scale, accessibility and commercial infrastructure of today’s systems are.
The practical threshold is delegation. Asking a tool for ten sketches leaves the artist with a visible selection task. Asking an agent to research references, choose a direction, generate assets, name files and publish the result transfers several judgments at once. Each step may be individually convenient while the chain as a whole becomes difficult to audit.
That is especially important where an agent can spend money, communicate under an artist’s name, retrieve private material or place work in public. A system can execute quickly without understanding the social meaning of a promise, the sensitivity of an archive or the consequences of a mistaken attribution.
A studio protocol for creative agents
Artists experimenting with agents can begin by drawing a boundary between recommendation and authority. Let the system propose suppliers, edits or captions, but require human approval before purchase, deletion, publication or direct communication. High-impact actions should remain reversible wherever possible.
Next, preserve a usable record. Keep the original brief, the sources an agent consulted, the tools it called, the versions it produced and the moment a person approved the result. A complete log is not automatically truthful, but it is far better than reconstructing a process after a dispute.
Access should also be limited to the smallest necessary set of files and services. A research agent does not need a studio’s master archive. A production assistant does not need permission to post publicly. Separate accounts, constrained folders and spending limits turn a vague promise of control into concrete safeguards.
Finally, decide how agency will be disclosed. If an autonomous system materially shaped the selection, sequence or behavior of a work, that fact may belong in the process documentation even when the final wall label remains concise. Transparency is not a confession; it is part of describing the medium accurately.
The artistic question is also a governance question
Agentic Aesthetics arrives at a useful moment because artists can test autonomy without accepting the technology industry’s default assumptions. Art can make delegation visible, stage conflicts between goals and expose the gap between a system’s apparent intention and its actual instructions.
The most productive question is not whether an agent is creative in the abstract. It is who defined its objective, what materials and permissions it received, which actions required consent, and who remains answerable when the result leaves the screen. For artists, authorship in the agentic era may depend as much on designing those boundaries as on producing the final form.
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.