Artist Molly Crabapple called generative AI the biggest art theft

Artist and journalist Molly Crabapple called what is happening around training models on other people's work the largest art theft in history in a column for Guardian. The text brings back to the center of the discussion not the quality of images, but a more uncomfortable question: who paid for the data on which the entire industry was built.

Where the conflict came from

The dispute did not begin with abstract fears, but with very concrete coincidences. In 2022–2023, artists began finding images on the internet suspiciously similar to their style; sometimes it was enough to enter the author's name in a prompt. AP wrote that Kelly McKernan, Carla Ortiz, and Sarah Andersen sued Stability AI, Midjourney, and DeviantArt, arguing that their work and millions of other images were used to train models without consent and without payment.

For artists, this is not an experiment with a new brush, but commercial repackaging of already created labor. The heart of the dispute is not only the generators themselves, but the data chain behind them. AP linked the boom in AI-generated images to the massive LAION dataset, which indexed vast amounts of publicly available pictures.

During hearings in the U.S. Senate in 2023, Ben Brooks, head of public policy direction at Stability AI, acknowledged that arrangements to pay artists did not exist at that time.

For critics, this is enough to call the current market model asymmetrical: technology companies get raw materials almost for free, while authors face competition from systems trained on their own portfolios.

How artists responded

The response was not only lawsuits. In May 2023, the Center for Artistic Inquiry and Reporting published an open letter calling for restricting the use of AI illustrations in media; it was supported by journalists, editors, authors, and artists. The letter states that image generators are trained on millions of protected works without the knowledge and consent of their creators, and the result is already displacing people from editorial assignments. Crabapple became one of the faces of this campaign, and her current column essentially continues the same line.

"This is effectively the largest art theft in history."

• Use of other people's work without consent
• Absence of payment and author attribution
• Emergence of cheap imitations of recognizable styles
• Replacement of editorial illustration with machine-generated imagery
• Disappearance of starter commissions for young artists

The price for the market

Crabapple's most painful argument is not related to aesthetics, but to economics. Illustration was already living on a thin margin, and generative services gave editorial offices and brands a super-cheap alternative that could be obtained in minutes. It is not star contracts that disappear first, but small commissions, on which young authors built portfolios and learned to work with clients.

If this market layer dissolves, the industry loses not only current income, but also future professionals. The conflict is also intensified by statements from tech companies themselves. In 2024, Mira Murati said that some creative work could disappear, and suggested that some of it may not have needed to exist in its previous form.

For artists, such formulations sound like an acknowledgment of substitution logic: first the platform learns from human labor, then that same labor is declared redundant. That is why the dispute around AI art has long gone beyond taste. This is now a question of who owns cultural raw materials and who gets paid when that raw material is turned into a product.

What this means

Crabapple's column is important not because it offers a new legal argument, but because it again makes visible the price of generative AI for the creative market. Until courts, publishers, and platforms agree on rules for licensing, consent, and compensation, every model improvement will be perceived by part of the industry as an accelerated form of appropriating other people's work. And the longer this gap persists, the harder it will be to convince authors that AI is a tool for collaboration, rather than a system of one-sided benefit.