Thomson Reuters v. Ross: the first AI fair use ruling explained

What did the court decide in Thomson Reuters v. Ross?

In 2025 a US federal district court ruled on fair use in a dispute between Thomson Reuters, owner of the Westlaw legal research service, and Ross Intelligence, a legal-research company. As widely reported, the court rejected the fair use defense for the specific use at issue: copying editorial content from Westlaw to help build a competing AI research tool. The decision was then sent up for appeal, so its final weight depends on what the appeals court does.

Read it as one data point, not a rule for every AI training question. The tool at issue was not a generative chatbot, the material was a publisher's proprietary editorial work, and the use competed with the original product. This is general information, not legal, tax or financial advice. We have not verified the current appeal status or the final outcome, and this page cites no court document. Confirm the current status of the case in the court's own docket and opinion, and with your counsel, before relying on any summary, including this one.

Why does a legal research case matter to business-data partners?

It matters because the content was business content, not consumer media. Westlaw's value comes from editors summarizing and organizing legal material for paying professional customers. That is the same kind of asset a services firm, software company or data publisher builds over years: structured, curated, commercially valuable work product.

Three practical lessons carry over to the companies you might introduce:

• Proprietary business content has commercial value that courts take seriously when it is copied without permission.
• Whether a use harms the original owner's market is a central question, which is why a working licensing market helps owners. The guide on why a functioning licensing market matters in AI fair use cases covers that thread.
• Developers who want clean inputs have a reason to prefer permissioned material, which is the supply side of what enterprise AI data licensing deals look like.

How does the case apply in common partner conversations?

You are not giving legal advice, but you will hear these reactions from owners, CFOs and advisors. Use the table to know what to check and where to point them.

What did the ruling not decide?

Several things stay open, and honest answers here build trust with your contacts:

1. It did not set one rule for generative AI. Other courts are weighing different facts, and results may differ.
2. It did not address internal business records such as email or support tickets, which are rarely public.
3. It did not tell any company that licensing is required or that unlicensed use is always unlawful.
4. It did not settle the appeal posture; later decisions may narrow, confirm or reverse it.

For the regulator's view of the broader training question, see the summary of the US Copyright Office report on generative AI training.

What does this mean for a referral partner?

Your job is not to interpret case law. It is to recognize a company with records it created, rights to license them and a sponsor who will talk. The case is useful as background when an owner asks why buyers pay for permissioned data at all.

AI developers, including the kind described in what is an AI lab, increasingly need real-work records to train and test agents, a point developed in the explainer on AI agent benchmarks and private business tasks. Rights-cleared records reduce the dispute risk buyers worry about, which is part of why chain of title matters in the process.

Why does ownership still come first?

Under 17 U.S.C. section 201, copyright vests initially in the author, an employer is the author of work made for hire, and exclusive rights can be transferred or licensed separately. In plain terms, a company can license specific rights in material it owns while keeping the rest. That is the structure a data license uses, and it is why the first screen for any company is whether it created the records and holds the rights, not what any court has said about someone else's content.

Anything created by clients, contractors without assignments or third parties needs separate consent. Raise that early, because it is the most common reason otherwise strong candidates stall.

What should you say if a contact asks about the case?

How are partner rewards connected?

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 is payable only after the buyer pays and SourceX receives its fee, and no reward is guaranteed. The reward is a share of SourceX's fee and is never deducted from what the company receives. Licensed professionals, such as lawyers, should check their own rules on referral fees and disclosure before registering.

Questions to take to your own counsel

• Does the current status of the Thomson Reuters v. Ross appeal change how we describe fair use to clients?
• Which of our client's records could be subject to third-party or client rights?
• Do our employee and customer agreements permit licensing this material?
• How should exclusivity for AI training be scoped in the agreement?

Confirm with your own counsel before acting.

Next step

Run one candidate through the company fit checker, a preliminary, non-binding screen that needs no contact details, and read how the process works end to end. If the company looks right, register as a partner and make the introduction.