> ## Content Index
> Fetch the complete content index at: https://wire.fourthweb.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# OpenAI Wins: Government Says Training AI on Copyrighted Work Is Fair Use
- URL: https://wire.fourthweb.ai/openai-wins-government-says-training-ai-on-copyrighted-work-is-fair-use/
- Published: 2026-09-02T17:09:06.000Z
- Updated: 2026-09-02T18:00:46.000Z
- Description: The U.S. just told every AI company in America that training on copyrighted data is probably fine, and every creator that their lawsuit strategy just got harder.
- Author: Travis Wright
- Tags: AI Agent Economy, AI Agents, AI Infrastructure, Compute Wars, OpenAI, Anthropic, China AI

**The U.S. just told every AI company in America that training on copyrighted data is probably fine, and every creator that their lawsuit strategy just got harder.**

### The Summary

- [The U.S. government filed an amicus brief supporting OpenAI's position](https://techcrunch.com/2026/09/02/u-s-government-sides-with-openai-on-issue-of-training-llms-on-copyrighted-material/?ref=wire.fourthweb.ai) that training LLMs on copyrighted material should be allowed under current law
- The brief frames AI leadership as a national security issue: ["The United States has a strong interest in continuing to develop a robust and competitive artificial intelligence industry that sets the standard for the practice and procedure of AI use globally"](https://techcrunch.com/2026/09/02/u-s-government-sides-with-openai-on-issue-of-training-llms-on-copyrighted-material/?ref=wire.fourthweb.ai)
- This isn't just legal cover for [OpenAI](https://wire.fourthweb.ai/tag/openai/). It's the government picking sides in the biggest IP battle of the agent economy.

### The Signal

The DOJ just handed OpenAI and every other foundation model builder a gift. In an amicus brief filed in ongoing copyright litigation, [the U.S. government argued that training AI models on copyrighted works serves national interests](https://techcrunch.com/2026/09/02/u-s-government-sides-with-openai-on-issue-of-training-llms-on-copyrighted-material/?ref=wire.fourthweb.ai). Not just economic interests. Global competitive interests. The kind of language usually reserved for semiconductor supply chains and rare earth minerals.

This matters because OpenAI, [Anthropic](https://wire.fourthweb.ai/tag/anthropic/), Google, and Meta have all faced lawsuits from authors, artists, and publishers arguing that scraping copyrighted material to train LLMs is theft. The New York Times sued. Sarah Silverman sued. Getty Images sued. The legal theory: if you wouldn't let a human read a million books without paying, why should a machine get to?

> "The United States has a strong interest in continuing to develop a robust and competitive artificial intelligence industry that sets the standard for the practice and procedure of AI use globally."

The government's position flips that framing. Training isn't consumption, it's transformation. And transformation at scale is how America stays ahead of China in the race to build the agent layer of the economy. The brief doesn't use the words "fair use" explicitly in the excerpt provided, but the logic points there: learning from copyrighted works to create something new has always been protected.

**What this means in practice:**

- Foundation model companies now have the executive branch backing their training practices
- Licensing deals with publishers become optional, not existential
- The cost structure of building new models just got dramatically cheaper if this position holds in court

The timing is everything. We're months away from models that can write full novels, generate feature films, and build entire codebases from scratch. If training those models required paying every copyright holder whose work touched the training set, only three companies on Earth could afford it. The government just said: we're not going to let that happen.

### The Implication

If you're building agents, this is a green light to train on everything you can legally access. If you're a creator counting on copyright law to protect you from AI competition, start thinking about what you make that can't be replicated by a model. The courts haven't ruled yet, but when the DOJ files a brief like this, it's signaling where the law is headed.

Watch how foundation model companies respond. Expect less apologizing, more scraping. And expect the next wave of litigation to focus not on training, but on output: when does a generated work cross the line from inspired by to copied from?

### Sources

[TechCrunch AI](https://techcrunch.com/2026/09/02/u-s-government-sides-with-openai-on-issue-of-training-llms-on-copyrighted-material/?ref=wire.fourthweb.ai)