Arthanyel Mistwing
LightNews LightNews
arthanyel.bsky.social
Arthanyel Mistwing
@arthanyel.bsky.social
160 followers 320 following 47 posts
Social Justice Paladin
Posts Replies Media Videos
Reposted by Arthanyel Mistwing
edzitron.com
There are 30 straight pages of Disney and Universal's copyrights juxtaposed with Midjourney outputs. Worse still, Disney and Universal have catalogued exactly how Midjourney trained their models, including a 2022 Forbes interview where they say they didn't ask for permission.
12. In response to the prompt, “Po character from the animated film Kung Fu Panda, full body view, screencap” Midjourney accessed the data about Universal’s Copyrighted Works that is stored by the Image Service and then reproduced, publicly displayed, and made available for download an image output that copies Universal’s Po character from the Kung Fu Panda franchise, as shown in this screenshot. permission, copied Plaintiffs’ Copyrighted Works to train and develop its Image Service. It also shows that Midjourney’s AI Image Service, through software programming and computer storage devices, embodies “copies” of those works, as that term is defined in Section 101 of the Copyright Act, in the AI software that powers the Image Service. On information and belief, and unbeknownst to Disney and Universal, Midjourney developed and trained its Image Service on copyrighted works, including Plaintiffs’ Copyrighted Works to help ensure that it is able to reproduce, publicly display, and distribute faithful, high-quality copies and derivative works of the works it trained on, including Plaintiffs’ Copyrighted Works. 115. Based on the facts and circumstantial evidence alleged in this Complaint and common knowledge about the process for creating a generative AI image service like Midjourney, Plaintiffs are informed and believe that Midjourney engaged in the following conduct to train its Image Service: a. First, Midjourney gathered the underlying works to be used to train the service. To do so, Midjourney downloaded from the internet, and other sources, content using tools variously described as bots, scrapers, streamrippers, video downloaders, and web crawlers. David Holz admitted that to collect the training data, Midjourney “pulls off all the data it can, all the text it can, all the images it can.”5 Mr. Holz elaborated saying that Midjourney, “grabs everything they can, they dump it in a huge file, and they kind of set it on fire to train some huge thing.” Admitting even more, in a September 2022 interview, Mr. Holz confirmed that Midjourney never sought any copyright content holders’ consent to copy and exploit their works.6 The data gathered and copied by Midjourney to use in training its Image Service includes Plaintiffs’ Copyrighted Works. b. Second, to prepare the data for ingestion, Midjourney “cleaned” the copies of the underlying works that were collected in the previous step through a filtering process and reformatted (e.g., converted them to a common technical format) the copies that were not filtered out in the cleaning process to train its Image Service. This step necessarily included creating more copies of the materials obtained in the gathering phase, such as a new copy of each reformatted item (including Plaintiffs’ Copyrighted Works). c. Third, Midjourney then used the collected and cleaned data and Copyrighted Works to “train” its Image Service. Although the specifics of this training phase have not been disclosed by Midjourney, and will be the subject of discovery in this action, Midjourney’s ability to repeatedly access the data stored in its software system and reproduce, publicly display, and distribute further copies of Plaintiffs’ Copyrighted Works for its subscribers demonstrates that Midjourney’s training of its generative AI model involved the fixation of copies of Plaintiffs’ Copyrighted Works in a tangible medium from which the work can be perceived, reproduced, or otherwise communicated with the aid of a machine or device. Midjourney’s process of training and making multiple copies of Plaintiffs’ Copyrighted Works was done without Plaintiffs’ approval or authorization. 116. In other words, Midjourney used software, servers, and other technology to store and fix data associated with Plaintiffs’ Copyrighted Works in such a manner that those works are thereby embodied in the model, from which Midjourney is then able to generate, reproduce, publicly display, and distribute unlimited “copies” of Plaintiffs’ works as defined by the Copyright Act
June 11, 2025 at 4:20 PM
Reposted by Arthanyel Mistwing
pressfreedomtracker.us
If you are a journalist — or know one — who has been attacked, seriously threatened, denied access by a public official or targeted with a legal order in the course of reporting, let us know. You can now submit tips in English or Spanish.

Email: tips@pressfreedomtracker.us
June 11, 2025 at 4:30 PM