ConcludedRan 1 dayin the edition since 4 Sep 20264 sourcesupdated 4 Sep 2026

DOJ brief on AI training and the NYT copyright suit

The gist

Latest: DOJ Backs OpenAI Fair-Use Argument in NYT Copyright Suit

The federal government's formal intervention signals the administration's position that AI training should be shielded from copyright liability at the training stage. The court must still rule on fair use case by case, but the DOJ framework, if adopted, would shift legal pressure from training toward data acquisition methods and specific outputs.

When stories like this move, I write about them in an email each morning.

One free email each morning, around 8 AM US Eastern.

The full picture

The U.S. Department of Justice filed a statement of interest in the OpenAI vs. New York Times copyright lawsuit, arguing that training large language models on copyrighted text qualifies as fair use. The DOJ described LLM training as 'transformative, spectacularly so,' on the grounds that models learn statistical relationships rather than reproduce content for the same purpose as the original. The administration also contended that training does not substitute for original works in the market, so AI-generated competition alone should not retroactively make prior training unlawful. The DOJ distinguished three separate acts, acquiring training data, training on it, and generating outputs, and focused its fair-use argument on the training stage alone, leaving open separate questions about data acquisition and specific outputs that reproduce protected passages. The filing is advisory rather than binding on the court.

How it developed

4 September 2026

Further reporting confirms DOJ brief supports OpenAI's fair-use position

First in the edition · 4 September 2026

3 September 2026in the 4 Sep edition

The U.S. Department of Justice filed a statement of interest on September 3 in the OpenAI vs. New York Times copyright lawsuit, arguing that training large language models on copyrighted text qualifies as fair use.

The DOJ called LLM training 'transformative, spectacularly so,' contending that models learn statistical relationships rather than reproduce content for its original purpose, and distinguished the training stage from separate questions about how data was acquired and whether outputs reproduce protected passages. The administration warned that rules making AI development harder in the United States 'threaten national security and give a competitive advantage to foreign adversaries,' adding that blanket licensing could disadvantage smaller U.S. firms. One commentator said the brief argues on national security grounds to invalidate copyright law and noted 'this is not how law works'.

First citedRohan Paul on X The Neuron Zvi's AI Roundups

Sources

Rohan Paul on X

The Neuron