This artist’s work has been shown at MoMA. Now it’s training AI

Over the course of his 50 years in the art world, Michael Hafftka’s figurative expressionist work has been exhibited at many of the world’s most prominent galleries. His paintings have hung at the Museum of Modern Art, the Metropolitan Museum of Art, and the Sauf Gallery in Paris. Now, his work is being presented in a more unusual place: on Hugging Face, the AI website.

The New York–born artist, now 72, has uploaded roughly half his oeuvre to the platform. He did it on his own initiative after researching Hugging Face and recognizing it as a gathering place for AI work. The move functions as both an artistic gesture and an archival one.

His path to AI is less surprising than it might seem. Hafftka has long gravitated toward emerging tools, from early experiments with computer-based art to more recent efforts with Web3. He was already plugged into the communities tracking generative AI as it began to accelerate, and followed that momentum in. He describes the project as “a new kind of catalogue raisonné,” a living record of his work, and a way to make it available for noncommercial AI experimentation.

“I decided that I could best use my work as a huge dataset, because so many artists are reluctant to collaborate with AI, but I like the idea of doing that,” says Hafftka. “That way I could actually wind up with a much better-trained model.”

Hafftka stands out as unusual in that space for embracing tech, and specifically AI. Most artists are opposed to the rise of AI, with some 61% of artists agreeing that AI is a threat to their livelihoods. Despite that, and because of the ease with which art can be made using simplistic prompts through tools like DALL-E, Google’s Nano Banana and Midjourney, the web is now flooded with AI-generated artwork.

While Hafftka is a tech booster, he’s also a pragmatist. “It’s inevitable anyway,” he tells Fast Company. In other words, if you can’t beat them, join them.

Still, Hafftka doesn’t see this as surrendering his future to AI. He argues that many artists have overstated their fears about the technology. “All of the fear that artists have about copyright usage rights, from actually making physical work, I think those fears are naïve,” he says. “We as a human being make something different from a machine, and I don’t really have the worry that I’ll be overtaken by machines.”

So far, the dataset—about 40.4 GB—has been accessed roughly 5,500 times, though Hafftka hasn’t yet seen anyone do anything with its contents.

Where others worry about their work being scraped, Hafftka is more concerned about how it’s used. Low-resolution versions of his paintings already circulate through museum collections online, and he assumes those are being pulled into training sets. “Those reproductions represent a small portion of my work”—though he plans to upload the rest in the coming months. “When that work is scraped from the internet and included in data sets, then I’m being done a disservice, because I feel all of my work is valuable and interesting, and I would rather that models training on my work be comprehensive.”

That position diverges sharply from artists suing or protesting AI companies. Hafftka’s concern is not that machines might learn from him, but that they might learn from him poorly. “Using one or two of his paintings is not going to reflect the brilliance or the interesting, unique aspects of John Singer Sargent,” he says.

He also situates AI within a longer arc of artistic change. “You can never stop technology,” he says. “What art is about is actually experimenting in order to create a unique and different vision.”

His confidence is rooted partly in the continued value of original artworks, which in his case sell for tens of thousands of dollars. “You make art for the world,” he says. “I sell original, and original cannot be duplicated by AI printing technology as it is today.”

He frames artistic influence less as theft and more as a continuum. Citing the late painter Jean-Michel Basquiat, whom he knew personally, Hafftka notes that imitation has always surrounded major artists without diminishing their work. “That’s how art progresses in the future,” he says. “Artists influence each other,” adding that “the idea of keeping it all locked down and afraid of future technologies is kind of paranoid.”

That makes Hafftka something of an outlier for now. But he doubts he’ll be alone for long. “I’m just the first to do this, because I have so much experience now with technology,” he says.

View the full article