Getty Images has launched legal proceedings against the creators of the AI art tool Stable Diffusion. Getty Images' lawsuit, filed this week in London's High Court, alleges that "Stability AI unlawfully copied and processed millions of copyrighted images" and used these images for its own commercial gain.

Rights holders of image libraries like Getty Images and the company will be most affected by AI image generation. Some, like Shutterstock (opens in new tab), have embraced the technology, while others have largely rejected AI art. Getty Images falls into the latter category. In September 2022, Getty Images banned the uploading and sale of AI images to protect itself from legal challenges (opens in new tab).

Getty CEO Craig Peters had previously stated that there were "unaddressed rights issues" and other concerns with AI-created artwork.

Little did we know at the time that the legal challenge would come directly from Getty itself.

"Getty Images believes that artificial intelligence has the potential to stimulate creative activity," the Getty Images statement (opens in new tab) reads. Getty Images has therefore offered licenses to key technology innovators for purposes related to the training of artificial intelligence systems in a manner that respects individual and intellectual property rights."

"Stability AI did not seek such licenses from Getty Images, but instead , We believe that it chose to ignore viable licensing options and long-standing legal protections in pursuit of its sole commercial interest."

"Stability AI has not sought a license from Getty Images.

It all comes down to how AI is trained, an important preliminary step in building a functioning artificial intelligence tool.

The art produced by Stable Diffusion and other tools like it is original, because it is created by the AI based on whatever prompts it is handed by the human user. However, the AI must be trained on millions of labeled images to learn what to create. The significance of the copyright of the images used for this training has been questioned many times.

Large datasets of such image-text pairs are often provided by other companies that collect and collate the data. These are huge datasets, with millions to billions of entries each. This means that it takes a huge amount of human-created source material to collect enough samples to fill an entire dataset, and each of those images has its own copyright position, ranging from public domain to strict copyright and terms of use.

One such dataset organization is LAION (opens in new tab), which provides the datasets used in Stable Diffusion.

LAION is non-profit and distributes its datasets freely. Basically, the datasets are collated from alt text and URLs of images scraped from the Internet; LAION does not own anything; LAION tries to assuage that in the first question of its FAQ (opens in new tab): "LAION's datasets are respect copyright laws?

To which the organization replies: "The LAION dataset is simply an Internet index.

There may be photos of you in LAION's dataset, but to comply with EU law we provide a GDPR deletion form.

LAION is actually getting its hands dirty by not directly storing or distributing the images. Getty Images' ire was directed at Stability AI, one of the well-known companies using LAION's datasets.

Stability AI states that the LAION datasets it uses for Stable Diffusion have been trained in compliance with German law; see the FAQ's "Copyright in the use of images generated by Stable Diffusion? ?", Stability AI states that "The area of AI-generated images and copyright is complex and varies from jurisdiction to jurisdiction.

A very vague answer indeed.

Stability AI at least states the source of the data - LAION is actually one of the more open image scraping organizations on the web, as is OpenAI (opens in new tab), which has developed popular AI tools like DALL-E and ChatGPT. Stability AI plans to allow artists to opt out of Stable Diffusion's image training (opens in new tab) in a future version, but that seems a little misguided. Stable Diffusion is a company that allows artists' work should not require the artist's permission to use the artist's work? [At this time, Stability AI has not responded to our legal action, but we have reached out to the company for comment.

Getty Images' CEO was interviewed by The Verge (opens in new tab) and further explained why the company took such action.

"We do not believe that this particular development of Stability's commercial offering is subject to fair dealing in the UK or fair use in the US," Peters said.

"Since the company has not lobbied Getty Images to use our or our contributors' material, we are taking action to protect our and our contributors' intellectual property rights.

Peters also confirmed that the charges against Stability AI include copyright infringement and violation of Getty Images' terms of use. He stated that the company aims to shape the new legal status quo regarding licensing and AI from this trial.

This status quo could be formative for the nascent generative AI tools industry. Copyright law has not yet caught up with training AI on the vast data sets of scraped images. In other words, legal actions over the next few years could shape the way we approach this topic for decades to come.

The lawsuit filed by Getty Images will likely set a legal precedent, whatever the outcome. [And the copyright debate surrounding both AI-generated images and AI tool training has only just begun.