As AI-generated art moves from novelty to exhibition wall, curators face a decision that traditional connoisseurship never prepared them for: an artwork can be aesthetically strong yet ethically compromised by how it was made. The core ethical considerations fall into four dimensions — data and consent, authorship and artistic labor, bias and representation, and transparency and disclosure — and each can be translated into concrete questions that make an exhibition or acquisition decision defensible rather than merely intuitive.

The Rise of AI Art and Its Ethical Imperatives for Curators

AI-generated work is no longer confined to online experiments; it is appearing in galleries and exhibitions, and with it comes a set of questions that reach far beyond whether the piece "works" visually. The scholars Johnathon Hall and Damian Schofield, writing in a 2025 study on the perceived value of human versus AI-generated art, note that ongoing ethical debates about AI in creative fields — including copyright, intellectual property, and the potential displacement of human artists — can influence both consumer perceptions and the art market itself.

For a curator, this changes the job description. The traditional evaluation of an artwork centered on originality, technique, provenance, and significance within an artistic lineage. An AI-generated piece adds a layer: the ethical profile of its production. A teaching resource developed by Stanford's CRAFT program frames the basic dilemmas bluntly — asking whether it is ethical for AI to use artists' work without permission or attribution, and how to debate dilemmas like cultural appropriation and unauthorized data use.

These are not abstract concerns for a later date. Because ethical questions can shape how audiences perceive and value AI art, as Hall and Schofield argue, a curator who ignores them risks mounting an exhibition whose rationale collapses under the first serious public challenge. What is needed is not a new aesthetics but a structured way of seeing the work — one that treats data sourcing, human contribution, representational bias, and disclosure as evaluation criteria alongside composition and concept.

Primary Ethical Challenges in AI-Generated Art

Copyright, Consent, and Data Provenance

The first cluster of concerns is about the raw material. Generative models learn from vast datasets of existing images, and the ethical flashpoint is whether the artists whose works fed those datasets ever consented to that use or received credit. Hall and Schofield's study records a strong desire among respondents for greater transparency in how AI models are trained, paired with calls for ethical standards ensuring "that artists' works are not exploited without due credit." The copyright question is genuinely unsettled law rather than a settled wrong — but the consent and attribution questions stand independently of any court's ruling, and a curator can ask them today. This article does not offer legal advice; the point is that provenance has become part of an artwork's meaning.

Authorship and Artistic Labor

The second debate concerns what, exactly, the work is. If a model generated the image, who is the author? The artist Craig Boehman offers a practitioner's counterpoint to the "AI replaces artists" framing: he uses a custom-trained AI chatbot to craft highly detailed prompts, generates multiple variations, refines the results, and then processes the chosen image further in professional editing software. In his account, the final piece is "a blend of AI-assisted generation and my personal artistic touch," with AI speeding the process "without replacing the artist's role."

Boehman's is one artist's practice, not a universal model — some AI works involve far less human intervention. But it illustrates the spectrum a curator must locate each work on: how much prompting, selecting, refining, and conceptual framing went into the final piece, and whether that labor is visible and honestly described. The question of originality, and how much human artistry the work actually contains, sits at the center of the debate over AI art's legitimacy.

Bias and Representation

The third challenge is structural. As Hall and Schofield observe, AI models trained on datasets that predominantly feature certain cultural styles or themes may produce artworks reflecting those biases, resulting in limited diversity in AI-generated art. A model that learned mostly from one tradition will reproduce that tradition's visual assumptions — and may embed stereotypes about people and cultures along the way. A curator evaluating an AI work is therefore also implicitly evaluating a dataset, asking whether the piece uncritically mirrors its training material or knowingly engages with, challenges, or exposes those biases.

Beyond the Human: A Decentered Ethics

A more forward-looking lens comes from the University of Melbourne's Centre for AI and Digital Ethics, whose Art, AI and Digital Ethics research program explores what it calls a "decentered ethics" of AI — an approach that moves past purely human-centric concerns to ask broader questions about technology's place in society and culture. The program is guided by two overarching research questions: what, specifically, artists and curators bring to the conversation about AI and digital ethics, and how the aesthetic encounters and emotional engagements they produce can support ethical feeling and deliberation. For curators, this perspective suggests that exhibitions are themselves ethical instruments — the way work is framed, contextualized, and encountered shapes public deliberation about AI, not just public taste in images.