Some artists train AI art generators to generate billions of copyrighted images used to recreate their own style, without compensating the artists or seeking their consent. We are starting a legal battle against the alleged theft.
A group of artists represented by Joseph Saveri Law Firm filed a US federal class action lawsuit in San Francisco against AI art companies Stability AI, Midjourney, and DeviantArt for alleged violations of the Digital Millennium Copyright Act and infringement of copyright. caused Publicity, and Illegal Competition.
Artists in action Sarah Andersen, Kelly McKernan, and Carla Ortiz said, “This blatant display of their rights before their professions are eliminated by computer programs powered solely by their hard work.” It seeks to end a colossal infringement,” the complaint filed in court.
With tools like Stability AI’s Stable Diffusion, Midjourney, or DeviantArt’s DreamUp generator, people can input phrases to create artwork that resembles a living artist. Since the mainstream advent of AI image synthesis last year, AI-generated artwork has been highly controversial among artists, sparking protests and culture wars on social media.
One notable company not included in the list of companies named in the complaint is the creator of the DALL-E image synthesis model that successfully rolled the ball into mainstream generative AI art in April 2022. is OpenAI. The content of the training dataset and some commercial licenses of the training data have been obtained from companies such as Shutterstock.
Despite the controversy surrounding Stable Diffusion, the Joesph Saveri Law Firm is well aware of legal action against generative AI, but the legality of AI image generator operation has not been tested in court. In November 2022, the same company filed a lawsuit against GitHub over his Copilot AI programming tool for alleged copyright infringement.
Weak discussion, ethical violations
AI analyst Alex Champendard said: advocated Claiming artists’ rights without outright dismissing AI technology, criticizing new lawsuits in several threads on Twitter, write in, “The attorney who filed this complaint is not credible based on its content and the way it was written. For this reason, this lawsuit may do more harm than good.” We believe that the defendant may be harmed. “whatever the company says to defend the companymyself used against them.”
Champandard noted that he noticed that the complaint contained several statements that may have misrepresented how AI image synthesis technology works. For example, in the fourth paragraph of Section I, it states, “When used to generate images from user prompts, Stable Diffusion uses training images to generate seemingly new images through a mathematical software process. These “new” images are entirely derivative works of the training images, the specific images that Stable Diffusion renders when assembling a specific output. Ultimately, this is just a complex collage tool. “
In another section attempting to explain how Latent Diffuse Image Synthesis works, plaintiffs erroneously refer to a trained AI model as having “a directory of billions of JPEG image files on a computer.” Compare and “The trained diffusion model is any of its training images.”
During the training process, Stable Diffusion extracted from a large library of millions of scraped images. Using this data, its neural network statistically “learned” how certain image styles look without storing exact copies of the images it has seen. However, in rare cases, images in the dataset may be overrepresented ( Mona Lisa), a kind of “overfitting‘ can occur, allowing Stable Diffusion to spit out a representation closer to the original image.
Ultimately, if trained properly, the latent diffusion model will always generate new images and will not create collages or duplicate existing work. This technical reality may undermine plaintiffs’ claims of copyright infringement. Image generators are, to our knowledge, an open issue with no clear legal precedent.
such as illegal competition (by duplicating an artist’s style and duplicating it using machines) and violation of publicity rights (by allowing people to request artwork “in the style” of an existing artist without permission). A few other points in the complaint, less technical, could be stepped in court.
Despite its problems, the lawsuit comes after a wave of outrage over lack of consent from artists who feel threatened by AI art generators. As it has admitted, the tech companies behind AI image synthesis have raked in intellectual property to train models without the consent of the artists. Even if they are finally found to comply with established case law regarding the over-collection of public data from the internet, they are already on trial in the courts of public opinion.
“Companies building large-scale models that rely on copyrighted data can get away with it if they do so personally.” murmured “But it is very difficult or impossible to do it openly and legally,” said Champagnedar.
If the case goes to trial, the court must sort out the difference between ethical violations and alleged legal violations. Plaintiffs want to prove that AI companies are commercially profitable and make rich profits by using copyrighted images. They sought substantial damages and permanent injunctive relief to stop the companies claiming infringement from committing further violations.
When asked for comment, Stability AI CEO Emad Mostaque said the company had not received information about the lawsuit as of this writing.
