OpenAI and Microsoft executives called their own products "largely substitutive" for news. It's in the court record now.

What happened

Reuters reported on September 17 that court filings made public that day in the New York Times copyright case quote OpenAI and Microsoft leaders describing their own products in terms publishers have been using for two years.

Two phrases are in the record: the products are "largely substitutive" for journalism, and they represent an "existential threat" to publishers. The New York Post and Yahoo Finance carried the same filings.

These are not analyst opinions or plaintiff characterizations. They are internal statements from the defendants, now public, in the case that will set the reference price for everyone else's archive.

Why those two words are the whole fight

Fair use in the US turns on four factors. Three of them — purpose, nature of the work, amount used — are argued in the abstract and both sides have decent material. The fourth is different. It asks whether the new use harms the market for the original.

"Largely substitutive" is not a rhetorical insult; it is close to the legal definition of the thing that defeats fair use on the fourth factor.

That is why this filing landed harder than another lawsuit headline. OpenAI and Microsoft have spent two years arguing their outputs are transformative — a new thing, not a replacement. A document in which their own people describe the product as a substitute for journalism cuts against that argument using the defendants' own vocabulary.

Reuters framed it exactly that way: the quotes could undercut the fair-use defense.

What this is not

It is not a ruling. No judge has weighed these quotes yet. Internal statements get context, get explained, get argued about scope. A vice president writing an urgent memo is not the company's legal position, and courts know that.

It is not a settlement signal. Nothing in the reporting says either company is moving to settle with the Times or anyone else.

It is not new money. Nobody's check gets bigger this week because of an unsealed document. If you were counting on litigation to fund your operation, the timeline is still years, and the plaintiffs are still large companies with legal departments — not independent publishers.

And it does not reverse the DOJ's position in the same case. That filing still exists. This is one more input into a fight with a lot of inputs.

What it actually changes for you

Leverage, not cash. Every licensing conversation happening right now — Google's contribution pilot, direct deals with OpenAI, the scraping conversations nobody announces — is priced against an expectation of how the copyright cases resolve.

When the defense gets weaker, the price of a voluntary deal goes up. Axios reported on September 16 that media executives are already refusing the terms currently on offer. This filing is the kind of thing that hardens that refusal.

The practical read: if you are being offered a licensing deal in the next quarter, the offer was priced before September 17. That is your argument for waiting or asking for more.

What to do about it

1. Do not sign a perpetual, all-archive license this quarter. The legal environment is moving against the buyers, and perpetual terms lock in a pre-September price.
2. If you have an offer on the table, ask for a shorter term — two years, renewable — instead of a bigger number. Term length is easier to win than rate.
3. Save the Reuters piece. It is the citation you use when a licensing rep tells you fair use makes your content free anyway.
4. Audit what you actually own. Freelance contracts without explicit AI-licensing rights are the gap that kills deals at signature.
5. Keep your crawler policy and your licensing policy separate decisions. Blocking a crawler is a distribution choice; licensing is a revenue choice. Conflating them costs you both.
6. Set a review for when the court rules on summary judgment. That is the date that changes prices, not this one.