PayAuthorscom

News

AI copyright news, graded by source

Court rulings, company filings and allegations in the copyright cases against AI companies and pirate libraries, over books, articles, music, images, video and code. Checked for new filings every three hours; items are posted after review, so some appear later. Get these by e-mail · Court filings, free to read

Each item is graded by its source: a court’s ruling or order; a company’s own filing about itself; a defendant’s filing arguing its side; a plaintiffs’ filing; the claims in a complaint (allegation); or press reporting.

What the labels mean

Get these updates by e-mail

For all news.

How the e-mails work

Uses the search, kind and company you set above (not the month or sort). As we post means within about 15 minutes of a new item; digests go out in the morning, U.S. Central time, weekly on Mondays, and only when something new matches. Updates start right away, and every e-mail has a one-click unsubscribe. If you tick Catch me up, the earlier items that match also come two a day, oldest first, each morning until you are up to date; you can stop that alone from any e-mail. Someone you invite gets one e-mail asking them to accept, and nothing more unless they do; we don’t tell them who invited them. Privacy.

Company filingApril 15, 2026Kadrey v. Meta Platforms, Inc. · N.D. Cal. 3:23-cv-03417

Meta's answer in Kadrey admits Llama 1 through 4 were trained on Books3 data, and that a script removed lines containing the word "copyright"

Answering the authors' amended complaint, Meta admits it trained Llama 1, 2 and 3 using data that was included in Books3, and that the Llama 4 versions released in April 2025 were also trained using Books3 data. It admits substantially all of the text of some of the plaintiffs' books appears in Books3, and that it used data from LibGen, Z-Library, Sci-Hub, Common Crawl and its own web crawler to train one or more Llama models. It admits it began a LibGen download in or around April 2023 by both direct download and BitTorrent, and that while it downloaded parts of datasets by BitTorrent "some data was uploaded." It admits a script was created to remove repetitive text from training data and that the program was designed to remove lines containing the word "copyright"; it says repetitive text is removed in part to reduce memorization risk. Meta denies infringement and notes the court's June 2025 summary judgment ruling in its favor on fair use for training. These are Meta's own statements in its pleading.

Meta's answer to the amended complaint (Dkt. 714), paras. 49, 86, 90, 92-93, 121 ↗

Loading the rest of the news…

What the labels mean

Every item says what kind of source it rests on. A court ruling or order is the court's own statement. A company filing is what a defendant says about itself. A government filing is a statement by a government agency that is not a party to the case, such as a statement of interest; it is the government's position, not a ruling. An allegation is what a complaint or motion claims and has not been decided. Reported means press reporting we have not confirmed from the primary document. Summaries are our own; follow the links for the source.

This page is general information, not legal advice. An allegation is not a finding. Whether any of this matters for a particular work or claim is a question for a lawyer.