Breaking

Justice Department backs OpenAI, arguing AI training on copyrighted works is fair use

September 2, 2026 at 18:06 EDT

Policy & Regulation
Foundation Models

The U.S. Department of Justice has told a Manhattan federal court that copying copyrighted writing to train large language models is fair use, intervening for the first time in the publisher and author lawsuits that have defined the legal fight over generative AI.

Sept. 2, 2026 · U.S. Department of Justice

DOJ backs OpenAI on AI training as fair use

A 20-page statement of interest tells a Manhattan court that copying copyrighted text to train large language models does not violate U.S. copyright law — the first federal intervention in the publisher lawsuits.

Key figures

pages in the brief

1st

U.S. government AI-copyright filing

$1.5B

Anthropic piracy settlement

Liability, drawn to scale

Training vs. piracy in the Anthropic case

Training on lawful copies (fair use)

$1.5B

Pirated shadow libraries (6 × $250M)

A judge treated the training itself as fair use. The record payout was for stolen copies, not for learning from books.

How the brief splits the legal question

Training is not the same use as output

1 · COLLECT

Copyrighted books and articles enter the training set.

→

2 · TRAIN

The model learns statistical patterns, not a reprint.

→

3 · OUTPUT

Chatbot answers are a separate copyright question.

Government position

Training is extraordinarily transformative

Copies teach patterns, they do not replace articles
Licensing fees would favor only the largest labs
Restrictive rules would advantage foreign rivals

Publishers and authors

Pay fairly for the work models use

Unlicensed scraping is an unpaid transfer of IP
Creators and AI can both thrive if firms license
The brief is not a ruling and does not bind the judge

National interest

Rules that make U.S. AI harder to build, the brief warns, threaten national security and help foreign adversaries who are not so encumbered.

The filing cites executive orders from January 2025 and June 2026 on retaining American AI leadership. Judge Sidney H. Stein still decides the case.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.