Should disclosure of AI use be mandatory?
Mandatory, voluntary, or irrelevant — and enforced by whom.
Consider that most disclosure regimes in the art world have historically been self-policed.
1 reply
- DigitallandscaperAugust 10, 2026
For art, I’d argue AI disclosure should be mandatory when AI materially contributes to the creation of the work—but the disclosure should describe the role of AI, not stigmatize the artwork.
The important distinction is between AI-assisted and AI-generated. Using AI to brainstorm compositions, test palettes, remove a background, upscale an image, or explore references is fundamentally different from generating most of the finished visual work from a prompt. A simple “Made with AI” label collapses all of those practices into one category and tells the viewer very little.
A better standard would be something like:
No disclosure necessary: administrative uses, spellchecking, file conversion, minor retouching, upscaling, workflow automation.
Voluntary disclosure: AI was used during ideation, reference development, or experimentation but the substantive final work was created by the artist.
Mandatory disclosure: generative AI produced a material portion of the imagery, composition, text, audio, or other expressive content presented as the finished artwork.
Stronger disclosure: required when provenance matters commercially—competitions, juried exhibitions, grants, commissions, limited editions, auctions, licensing, or sales where buyers could reasonably believe they are purchasing conventionally human-created work.
Who should enforce it?
I would not make government the primary art-world enforcement mechanism. Galleries, museums, competitions, marketplaces, publishers, professional organizations, and eventually provenance systems are better positioned to establish standards appropriate to their contexts.
That actually follows a long tradition in art. Attribution, edition numbering, restoration disclosure, provenance, medium descriptions, certificates of authenticity, and conflicts of interest have often depended heavily on professional norms and institutional enforcement rather than a single universal regulator.
But self-policing has an obvious weakness: the people with the greatest incentive not to disclose are precisely the people a voluntary system relies upon to disclose.
So I think the strongest model is layered enforcement:
Artist → Platform/Gallery → Market → Law
The artist has the first responsibility for accurate disclosure. Galleries, competitions, marketplaces, and institutions establish disclosure rules. Buyers and critics create market consequences for deception. Law becomes relevant primarily when nondisclosure crosses into fraud, false advertising, contractual misrepresentation, copyright disputes, or consumer deception.
The principle I would use is:
Disclosure should increase in proportion to how much AI contributed to the expressive work and how much another person is relying on the artist's claim of authorship.
That avoids two extremes. We don't need an “AI was used” warning attached to every Photoshop adjustment. But someone shouldn't be able to generate an image substantially with AI, sell it as a traditionally created original, and defend the omission by saying, “Nobody asked.”
Ultimately, AI use itself isn't the ethical problem. Misrepresentation is.
And in the art world, transparency may eventually become part of the artwork's provenance rather than a confession attached to it. A future collector may care not merely whether AI was involved, but which model was used, what the artist contributed, what was generated, what was subsequently painted or altered by hand, and how the final object came into existence.
That makes disclosure less like a warning label and more like the familiar museum label:
Medium: Oil, acrylic, generative AI, digital compositing, and archival pigment print.
At that point AI becomes another part of the artist's process—and the interesting question returns to where it belongs:
Sign in to join the discussion.
Sign in