US Backs AI Training on Copyrighted Work: What Creators Need to Know

What changed: the US government entered the AI copyright fight

Confirmed fact: the US Justice Department filed a statement of interest in the copyright case brought by the New York Times and other publishers against OpenAI and Microsoft. The filing supports OpenAI's position that using copyrighted texts to train large language models is generally protected by the US fair-use doctrine.

Reuters reports that this appears to be the first time the US government has formally weighed in on the current wave of AI-training copyright cases. The government argued that training is highly transformative and that restricting it could slow scientific progress, economic growth and national-security work.

Just as importantly, the filing is advisory. The judge—not the government—will decide the case. US fair use is assessed case by case, and courts have not yet produced one settled rule covering every model, dataset or type of creative work.

The New York Times rejects the government's position and argues that AI companies should pay fairly for the content used to build their products. That dispute remains unresolved.

Why it matters: the rules behind the AI economy are still being written

Confirmed fact: major AI companies including OpenAI, Anthropic and Meta face copyright claims from authors, publishers and other rights holders. The decisions in these cases could influence what material developers can use, whether licences are required and how creators are compensated.

Our analysis: this matters to small creators because the value of original work is being tested from two directions. That work can help train products worth billions, while AI-generated answers and media may reduce the traffic, attention or sales reaching the original source.

The opposite pressure is also real. Affordable AI tools give solo businesses access to research, writing, design and automation capabilities that once required a team. A legal system that makes model development far more expensive could concentrate the best tools inside the largest companies.

The core question is therefore not simply whether AI is good or bad. It is how innovation can continue without making human creators invisible, uncompensated or unable to sustain the work that future systems depend on.

The opportunity: original, provable work becomes more valuable

Creators cannot control every training dataset, but they can strengthen the business built around their work. Clear ownership records, recognisable expertise and a direct relationship with an audience are harder to replace than anonymous, interchangeable content.

There may also be growing opportunities around licensed content, specialist datasets and verified human expertise. Businesses that can prove where material came from—and that they have permission to use it—will be more useful to clients who want AI-assisted content without unnecessary rights risk.

For creators using AI, the sensible advantage is speed with accountability: use tools to organise ideas, test formats and reduce repetitive work, then add experience, evidence and a point of view that belongs to you.

• Keep dated source files for valuable original work
• Build an email list or customer base you control
• Offer licensed assets or specialist knowledge with clear terms
• Use AI to assist production without copying a living creator's style
• Make human expertise visible in the finished content

The risk: a headline can be mistaken for permission

The biggest immediate risk is misunderstanding the news. The US government's filing does not give every business permission to copy articles, images, books or videos. Fair use is a specific US legal defence, not a universal exemption, and the underlying court case is still active.

UK businesses should be especially careful. The UK government's March 2026 copyright report says British law does not have the same broad fair-use principle as the United States and generally provides stronger protection for rights holders. Where material was created, copied, trained on or published can therefore matter.

A second risk sits inside everyday AI use. A generated result can reproduce distinctive wording, visual elements or false claims even when the user did not request them. Publishing that result under your name still exposes your reputation and business to the consequences.

This article explains a developing news story and is not legal advice. For high-value commercial work or a suspected infringement, speak to a qualified copyright professional in the relevant country.

One practical action: build a simple rights record today

Choose one valuable product, article, image collection or course that you own. Create a basic rights record showing when it was made, who created each part, where the original files are stored and what licences apply to any fonts, photographs, music or other third-party material.

Then review the terms of the AI tools used in that workflow. Check what the service says about uploaded content, model training, ownership and commercial use. Save the relevant policy link and the date you checked it alongside the project files.

This will not stop every misuse, but it gives you cleaner evidence of ownership, helps you avoid publishing material you do not control and makes future licensing conversations much easier.

• Save the earliest editable source files
• Record the creator and creation date
• List every third-party licence
• Check the AI tool's current data and ownership terms
• Store the record with the finished product