A 12 Million Image Lawsuit Puts AI Training Under Scrutiny

Getty Images sued Stability AI in the United States, alleging that it copied twelve million images without permission or compensation to train Stable Diffusion. The dispute raises questions about image licensing, copyright and whether AI training can rely on material gathered from online sources.

WTF Index TERMINATOR
◄ Terminator 2 Idiocracy 1 ►

The allegations raise concerns about unlicensed use of creators’ work, though the story centers on a legal dispute rather than AI harm or loss of human skills.

A 12 Million Image Lawsuit Puts AI Training Under Scrutiny

Getty Images has taken Stability AI to court in the United States, alleging the company copied twelve million images without permission or compensation to train Stable Diffusion. The case puts a commercial dispute over image use at the center of wider questions about how generative AI systems are built and what rights apply to the material they process.

Getty alleges unlicensed copying

In its complaint, Getty Images says Stability AI copied and processed copyrighted images, along with metadata, from its platform. The company says the material was used to train Stable Diffusion, an AI image generator, and argues that scraping its platform violated its terms of use.

Getty Images CEO Craig Peters said the company believes Stability AI used others’ intellectual property for its own financial gain. He described the issue as whether generative models must address the intellectual property rights of others, and said Getty was seeking clarity through its legal action.

The updated account says the lawsuit was filed in the United States District Court in Delaware. Getty alleges that Stability AI used links from LAION to create copies of images in its library. The complaint says the system can sometimes produce images that are highly similar and derivative of Getty’s collection.

The dispute also concerns licensing

Getty Images built its business around collecting and licensing images online. The growth of AI image generators raises a commercial challenge for that model, while also creating a possible new market: licenses for companies that want to use images in AI training.

Getty says Stability AI did not seek a license and chose to ignore available licensing options and legal protections. That is the company’s position in the dispute; the source article does not report a court ruling on whether the alleged copying infringed copyright.

AI copyright specialist Andres Guadamuz suggested the lawsuit could put pressure on Stability AI to pay royalties for images processed during training. He also raised the possibility that Getty could seek a more commercially favorable role as AI develops, including by supplying licensed images for training.

What the case may clarify

Getty is likely to argue that the use of its images amounts to infringement. Stability AI, according to the article’s account of the legal dispute, would likely argue fair use. The Delaware court has few fair use cases, Guadamuz said, leaving uncertainty over how the dispute might be resolved.

Copyright attorney Aaron Moss, cited via The Verge, suggested hearings could take several years. The article also noted that both companies might develop more advanced AI systems in the meantime and could settle the copyright issues out of court. Those are possibilities, not reported outcomes.

The case is part of a broader argument about who can use creative work to train generative systems. Three U.S. artists—Sarah Andersen, Kelly McKernan and Karla Ortiz—were also pursuing a class action against Stability AI, Midjourney and DeviantArt. Getty’s lawsuit gives the licensing dispute a major image platform’s perspective alongside artists’ claims.

Licensing and AI’s commercial future

Getty Images had already opposed AI-generated images in its library, citing unresolved copyright questions when it banned them the previous September. The source article describes a tension: AI-generated images may compete with a business built on licensing photographs, but image libraries could also offer licensed material for AI training.

Stability AI had announced an intention to use datasets containing licensed material in the future. That could point toward licensing as one way to address concerns about training data, though the source does not say whether Getty and Stability AI reached an agreement.

At the time of the article’s update, Getty had filed its U.S. case, while a separate UK lawsuit had not yet been served, according to a Getty Images spokesperson. The original account said legal proceedings with Getty Images would take place in London, and that Stability AI had not yet commented on the allegations. The competing claims and the role of licensing remained unresolved.