Imagen is an artificial intelligence (AI) software product created, maintained, and sold by Google. It is a text-to-image diffusion model—one that takes as input a short text description of an image (also known as a text prompt) and then uses a machine-learning technique called diffusion to generate an image in response to the prompt. 

Rather than being programmed in the traditional way—by human programmers writing code—a diffusion model is trained by copying an enormous quantity of digital images with associated text captions, extracting protected expression from these works, and transforming that protected expression into a large set of numbers called weights that are stored within the model. These weights are entirely and uniquely derived from the protected expression in the training dataset. Whenever a diffusion model generates an image in response to a user prompt, it is performing a computation that relies on these stored weights, with the goal of imitating the protected expression ingested from the training dataset. 

Training a model first requires amassing a huge collection of data, called a dataset. The AI models at issue in this litigation were trained on datasets containing millions of images paired with descriptive captions. Each image–caption pair is called a training image. During Imagen’s training, the training images in the dataset were directly copied in full and then completely ingested by Imagen, meaning that protected expression from every training image entered it. As Imagen copied and ingested billions of training images, it progressively developed the ability to generate outputs that mimic the protected expression copied from the dataset. 

Plaintiffs and the class are visual artists. They own registered copyrights in certain training images that Google has admitted copying to train Imagen. Plaintiffs and class members never authorized Google to use their copyrighted works as training material. 

These copyrighted training images were copied multiple times by Google during Imagen’s training process. Because Imagen contains weights that represent a transformation of the protected expression in the training dataset, Imagen is itself an infringing derivative work. Alphabet, as the corporate parent of Google, also commercially benefits from these acts of massive copyright infringement. 

 Blog Pictures (Blog Banner) (1) copy-1


On April 26, 2024, the firm and co-counsel filed a lawsuit on behalf of plaintiff and class-member visual artists who own registered copyrights in certain images that were included in training Imagen. The case, Zhang v. Google LLC and Alphabet Inc., in the United States District Court for the Northern District of California, seeks damages, as well as destruction or other reasonable disposition of works that defendants made or used in violation of the exclusive rights of plaintiffs and the class.  

"Artificial intelligence is dramatically changing every aspect of modern life and putting the livelihood of visual artists such as our clients at risk. Their rights must be recognized and protected against unlawful theft and fraud," said firm founder, Joseph Saveri. “The training of Imagen infringes our clients’ rights. This case, like several others recently filed by our firm, is part of a larger fight for preserving ownership rights for all artists and other creators."  


If you wish to inform us of any unfair business practice, antitrust or competition issue, or comment on one of our cases, please use the form below. There is no cost or obligation for our review of your case. We agree to protect your name and all confidential information you submit against disclosure, publication, or unauthorized use to the full extent under the law. Please note that completion of this form does not contractually obligate our firm to represent you. We can only represent you if both you and our firm agree, in writing, that we will serve as your attorney. Please read our disclaimer.