New York Times v. OpenAI: where the case stands and why licensing matters

The short answer, with the hedge

The New York Times Company filed a copyright suit against OpenAI and Microsoft in federal court in Manhattan in December 2023. It claims the companies copied Times journalism without permission to train AI models, and that the resulting products can reproduce or summarize that journalism, harming the paper's subscription and licensing business. The defendants argue their use is lawful, chiefly under the doctrine of fair use.

What the case means depends on rulings that keep arriving. This page was reviewed in October 2026 and deliberately does not restate rulings we could not verify against court records, so check the docket before relying on any summary, including this one. The question underneath does not change: when a market exists for licensing content to AI developers, how much does unlicensed use harm that market?

What the law says

Two pieces of US law frame the dispute for anyone thinking about data licensing.

• Fair use is decided case by case. US copyright law lets courts excuse some unlicensed copying as fair use, weighing factors that include the purpose of the use and its effect on the market for the original work. The US Copyright Office's report on generative AI training, released as a pre-publication version in May 2025 as Part 3 of its AI series, examines how fair use may apply to training and how practical licensing approaches are (Copyright Office AI initiative). It is an agency analysis, not law, and courts are not bound by it.
• Workplace records usually belong to the employer. The Copyright Act defines a work made for hire to include a work prepared by an employee within the scope of employment (17 U.S.C. 101). That is why a company, rather than its individual staff, typically holds the rights in its internal documents, subject to contracts with clients and vendors.

This is general information, not legal, tax or financial advice. Confirm with your own counsel, tax adviser or professional body before acting.

Why licensing markets matter to the fair use argument

The market-effect factor is where licensing enters. Rights holders argue that a functioning market for AI training licenses shows there is something to harm: if developers pay some publishers, taking from others without paying displaces a sale. Defendants respond that training serves a different, transformative purpose, and that a market cannot be created simply by demanding payment for a use that is otherwise fair. Courts weigh these arguments on the facts of each case.

The public record shows such a market forming. Dated examples:

Litigation and licensing are not opposites. The Times has since been reported to license content to another AI developer, covered in the reported Amazon–New York Times licensing deal, and each new agreement adds to the record that a market exists.

The case also reaches beyond training. Claims about models reproducing articles, and about answers that draw on current reporting, touch retrieval and display as well; training rights and retrieval rights are different grants, as the training vs RAG license comparison shows.

How the case applies in common partner situations

Disclosure and consent good practice

Whatever the court decides, buyers of licensed data want a clean chain of rights. Companies that license records through SourceX keep ownership, agree redaction and de-identification rules before any work begins, and deliver only after an executed agreement and their own authorization. Partners never export, upload or describe confidential records; they make the introduction and share basic fit information. The page on the Bartz v. Anthropic settlement explains what another AI copyright case means for licensed data.

How to check the case status yourself

1. Search PACER, the federal judiciary's records system, or a free docket mirror for The New York Times Company v. Microsoft Corporation in the Southern District of New York.
2. Check whether the case has been consolidated or coordinated with other copyright suits against OpenAI, and which docket now carries the filings.
3. Read the orders themselves, especially opinions on motions to dismiss, discovery disputes and summary judgment, rather than headlines.
4. Note the date of the latest order you read, and recheck before quoting the status to anyone.
5. Ask counsel what a given ruling means for your situation, since many rulings at this stage decide procedure rather than the final question.

Questions to ask your counsel

• Does any ruling in the case affect a company licensing records it created and never published?
• Which of our records contain material owned by clients, vendors or individuals?
• Do our privacy notices, client contracts and employee policies permit licensing for AI training?
• What warranties and indemnities would a buyer ask for, and can we support them?
• If the law shifts during the license term, what does the agreement say?

What it means for referral partners

Do not offer clients predictions about the case or opinions on fair use; point them to their counsel. The practical distinction is simpler: the dispute concerns published content allegedly used without a license, while a SourceX license covers records the company itself created, under terms it negotiates and signs. For a first screen, use the company fit checker, which is preliminary and non-binding; the how it works page shows each step from introduction to payment.

Partners earn 25% of the eligible platform fees SourceX actually collects from the referred company's licensing deals, capped at $100,000 per referred company; the reward becomes payable only after the buyer pays and SourceX receives its fee, and no reward is guaranteed.

Next step

If you advise companies with deep, company-created records, register as a partner and introduce the ones that pass a first screen. Background reading on how enterprise AI data licensing deals work helps in those conversations.