Art and generative AI: showing artists who create with artificial intelligence

01A rapidly recomposing artistic landscape

The use of artificial intelligence in art predates the explosion of mainstream generative models. From the 2010s, artists such as Refik Anadol, Holly Herndon, Trevor Paglen and Hito Steyerl were exploring the intersections between machine learning, big data and visual creation. Refik Anadol, whose immersive installations based on data processed by neural networks have been shown at MoMA, Centre Pompidou and the Venice Biennale, demonstrated that AI could be the medium for spectacular and emotionally powerful work. The Paris-based collective Obvious attracted media attention in 2018 when their portrait generated by a generative adversarial network (GAN) was sold at Christie's, sparking a debate that continues.

The current generation of artists working with AI is considerably more diverse and more sophisticated. Some use AI as an image-generation tool that they then rework manually, in a process akin to that of the photographer who selects, crops and retouches their shots. Others train their own models on specific data sets, creating bespoke generative systems that resemble no commercially available tool. Still others interrogate algorithmic bias, the intellectual property of training data, or the ethical implications of delegating creative decisions to the machine. The dealer entering this territory must map this diversity of practice to avoid reducing art and AI to a monolithic category.

The speed of evolution of this landscape is itself a challenge for the dealer. Tools change every six months, legal and ethical debates shift constantly, and institutional positions reconfigure with every exhibition and public statement. The dealer must maintain an active watch, attend specialised conferences, read academic publications and engage with artists to stay current in a domain that knows no stasis.

02Why collectors are paying attention

Collector interest in art generated or assisted by AI rests on several converging factors. The first is fascination with a technology that is redefining the boundaries of human creativity. The most informed collectors, often from the technology world, see in these works a mirror of their own professional and intellectual environment. They perceive an art form that engages with the issues of its era more directly than traditional painting or sculpture.

The second factor is speculative potential. The AI art market is still young, positions are not fixed, and early collectors hope to benefit from the future appreciation of artists whose prices remain accessible. This dynamic recalls the early days of the photography or video art market: a period of scepticism followed by progressive institutional recognition that drives prices upward.

The third factor is conceptual depth. Works created with AI raise profound philosophical questions about the nature of creativity, the artist's place in the creative process and the very definition of the work of art. These questions attract collectors who seek not merely a decorative object but an object of reflection, a piece that sparks conversation and stimulates thought. The dealer who can articulate this intellectual dimension in their mediation discourse adds a layer of value that justifies prices and strengthens collector loyalty.

03How the dealer evaluates artistic quality

The main challenge for the dealer is distinguishing the artist who uses AI with intention and sophistication from the mere user of a generative tool. This distinction is crucial for the gallery's credibility and collector confidence. Several criteria help make this judgment.

Artistic intentionality is the first criterion. The artist working with AI must be able to articulate clearly what the technology brings to their practice, why it is necessary to their purpose and how their use of it differs from simply issuing a command to an image generator. The dealer should pose these questions during the studio visit and assess the depth and coherence of the answers.

The creative process is the second criterion. An artist who merely types a prompt into Midjourney and selects from the results does not have the same standing as one who trains their own models, intervenes manually in the output, combines AI with other media or conceives installations in which AI is one element among others. The complexity and singularity of the process are indicators of the artistic value of the practice.

Programmatic coherence is the third criterion. The artist using AI must be engaged in an identifiable research programme, with visible evolution from one series to the next, recurring themes and an ambition that exceeds the production of attractive images. The dealer evaluates this coherence by examining the artist's trajectory, previous exhibitions, writings and public statements. An artist whose AI work is embedded in a long research trajectory, with references to art history and contemporary debates about technology, inspires more confidence than one who discovered AI last week.

04Legal and ethical questions

Showing and selling works created with AI raises legal questions the dealer cannot ignore. Copyright is the most debated issue. In most European jurisdictions, copyright protects works created by a human author. An image produced entirely by a machine, without significant human intervention, may not benefit from this protection. The dealer must ensure that the artist contributes sufficient creative input for the work to be protectable, and document this contribution to demonstrate it in case of dispute.

The question of training data is another concern. Generative models are trained on image corpora that often include copyrighted works. Artists and photographers have filed lawsuits against the developers of generative models, arguing that the use of their works as training data constitutes an infringement. The dealer must stay informed about the state of the law and ongoing proceedings, and be able to reassure collectors about the legitimacy of the works on offer.

The ethical dimension is inseparable from the legal one. Some artists refuse to use models trained on unconsented works and develop their own systems from ethically collected data. Holly Herndon, for instance, developed with Mat Dryhurst an AI system trained exclusively on the voices of willing contributors. The dealer who highlights this ethical dimension in presenting the works adds a layer of value that reassures collectors sensitive to these issues and positions the gallery on the right side of a debate that is only intensifying.

05Exhibiting AI art: scenographic challenges

Exhibiting works created with AI poses specific scenographic challenges. If the work takes the form of a photographic print or a manually reworked painting, classical hanging is adequate. But many AI-related works are digital installations, projections, interactive works or real-time generative systems that require adapted technical environments. The Unsupervised exhibition by Refik Anadol at MoMA in 2022-2023, which presented a digital work in perpetual transformation drawing on the museum's collection data, showed that such works can captivate a broad public when the scenography matches.

The dealer must anticipate these technical needs and invest in the necessary infrastructure: high-resolution screens, projectors, computing systems, network cabling. Pace Gallery, which opened a department dedicated to digital art and new technologies, demonstrated that this investment is commercially viable and attracts new clientele. Galerie Konig in Berlin regularly programmes digital art exhibitions, showing that a contemporary space can accommodate these works without abandoning its architectural identity. The cost of this infrastructure can be amortised by making it available to other galleries or institutions for one-off events.

The dealer must also think about the documentation of the work. For a real-time generative work that does not exist in a fixed form, the question of what the collector acquires must be clarified contractually. Is it the source code, access to the algorithm, a print generated at a specific moment, a complete installation with hardware? These questions, which have no universal answer, must be resolved case by case in dialogue with the artist.

06Positioning the gallery in a forming market

The AI art market is still taking shape, which means positions have not yet been allocated. The dealer who commits today can build a reference position before the market consolidates and the places are taken. Fairs such as Art Basel, Frieze and FIAC programme sections dedicated to new technologies and digital art, offering galleries presenting these artists visibility among international collectors.

The dealer should, however, avoid positioning exclusively in this segment. A gallery's credibility rests on the overall quality of its programme, and integrating artists working with AI into a broader programme that includes traditional media strengthens the legitimacy of digital works by placing them in dialogue with art history. A programme that mixes painting, sculpture, photography and digital art suggests the dealer selects artists for the quality of their work, regardless of medium, rather than riding a trend.

For galleries partnering with Artedusa, presenting artists who use artificial intelligence offers an opportunity to reach an audience of tech-savvy, international collectors open to contemporary forms of creation, while benefiting from a presentation framework that highlights the artistic dimension of these works beyond their technological component.