If you're developing an LLM in the United States and need loads of data, the government has made a legal filing that claims copyright violations by AI companies are fine, as long as it means moving faster than perceived adversaries.
I don't like to start my pieces off with a quote, but I feel like this colors everything:
"Rules of law that make it significantly more difficult to develop a robust AI industry in the United States therefore threaten national security and give a competitive advantage to foreign adversaries who are not so encumbered," is the quote that made me break my editor's rules.
That's a statement from an official filing from the United States Associate Attorney General Stanley Woodward Jr, in an amicus curae brief.
First, it should come as no surprise that OpenAI was the first to make the claim in March 2025. That statement arrived only days after the US government blacklisted Anthropic, then OpenAI took Anthropic's spot as the government and military's AI of choice.
Fast forward a year and a copyright lawsuit between OpenAI and The New York Times prompts the US government to file this opinion with the courts. The one signing the filing is a Trump appointee, placed in the role after representing January 6 Capitol attackers and Trump's valet in the classified documents case.
Woodward's arguments are nearly a copy and paste of OpenAI's original complaints about AI training and not coming from a place of legal authority. Everything from "IP theft is necessary because our adversaries won't honor copyright" to "it's too hard to regulate this, so just don't" is present here.
To make things even less viable, is OpenAI's involvement in the US military that was announced just a day before. Don't worry, Grok is also being used by the Department of Defense, in case you were wondering.
While this is an official filing from the US government, it is still just an opinion the court can consider and reject. It speaks for itself, so here are some more snippets.
"An erroneous fair use ruling would hamper competition in the market for LLMs, because only the largest technology companies might have the capital necessary to pay licensing fees."
That's a wonderful sentiment, except OpenAI, Anthropic, and Google already own the market. In fact, these companies have repeatedly complained about smaller AI startups "stealing" their frontier models.
Under the same logic, that should be allowed. And also following that logic, since OpenAI isn't making an iPhone clone, it's okay if OpenAI steals intellectual property from Apple.
"At a minimum, the creative possibilities and public benefits that LLM training advances far outweigh any competitive harm (even assuming such harm is cognizable). The entire reason that AI models are rapidly reshaping the economy and national security is that they help people, including those working in creative fields, make things and get things done."
This government filing restates the argument, almost verbatim, that OpenAI and others have been arguing for years. The filing makes the erroneous claim that AI models are reshaping the economy positively, and enabling users.
The reality has been that people are losing jobs, and money is being funneled towards the AI companies instead. As it turns out, outside of some disciplines like coding or research, this is a net-negative for the US economy.
Realistically, this argument continues to be a sign that AI companies are flailing against the eventual reality that this technology has its limits. AI isn't some endless pool of potential profitability, it's just really good analytical software that can hallucinate and get things terribly wrong.
OpenAI is also aware it's on borrowed time. It may never be profitable, and at best, it has until 2028 to get massive cash infusions through something like profitable hardware or face running out of money.
"For all these reasons, the United States has a strong interest in this Court rejecting any argument that training LLMs on copyrighted texts violates copyright law."
The "United States" is doing a lot of heavy lifting here. It would certainly benefit companies like OpenAI, which are being sued, but does little to address the problems caused by legal theft of copyrighted material.
"Any effect on the market for the copyrighted works cannot overcome the significant world-changing value of the training use's transformative purpose."
The whole argument here is that OpenAI is consuming this information that it did not license, and regurgitating it as an LLM output, so no original content remains. Except, time and time again, people are able to reproduce content consumed by the LLM training.
And, we've seen vast sections of our own copy plagiarized directly by AI, following a scrape, without citation. That is a blatant violation of "fair use" that the feds seem to be ignoring in their filing.
New York Times spokesman Graham James provided a statement on the matter:
"The Administration is siding with a handful of trillion-dollar Al companies at the expense of the countless American creators whose work they stole. Both Al and creators can thrive - Al companies simply need to pay fairly for the content that makes their products possible, as copyright law requires.
The Administration's proposal to let companies take that content without permission or compensation would undermine the sustainability of the human-created content that a healthy society depends on, and which Al needs to function."
Hopefully the court can make the right call, even with pressure from multiple AI companies and the Trump administration.
The case started in December 2023 and hasn't moved much since. As of September 2, 2026, the court requests any amicus briefs be filed by October 16, 2026.
