Thomson Reuters

What the company says it is

Thomson Reuters describes itself as an AI and technology company that provides professionals with trusted content and workflow automation^[1]. That phrasing is a fair guide to how the organisation presents itself: not simply as a publisher of information, but as a supplier of tools that sit inside other people's working day, in law, tax, accounting, government and corporate settings.

A long content base

The company says it draws on content the world has trusted for over 150 years^[1].

Labs and the institute

Two of the company's named units show how it organises this work. Thomson Reuters Labs is a global network of labs that explore new business opportunities and build experiments and proofs of concept together with customers and partners^[1]. The Thomson Reuters Institute, separately, brings together people from the legal, corporate, tax and accounting, and government communities^[1] — a convening role rather than a product one.

Westlaw

Among the assets Thomson Reuters owns is Westlaw, a legal research database. Westlaw includes headnotes and a Key Numbering System^[2]. These are not simply copies of court opinions. The headnotes at issue in the litigation below are editorial content that Thomson Reuters writes to summarise and organise judicial opinions^[2]. In practice, that editorial layer is what makes a large body of case law navigable: a reader can find a passage and see, in the company's own words, what point of law it stands for, and the numbering system groups related points together.

The dispute with Ross Intelligence

The company sued Ross Intelligence in 2020 over Ross's use of memos derived from Westlaw headnotes to train an AI legal search tool^[2]. The background matters: Ross first asked to license Thomson Reuters content to train its search engine, and Thomson Reuters refused the request^[2]. The training material at issue therefore came from the editorial headnotes rather than from the underlying judicial opinions alone.

The case was heard by a federal judge in Delaware^[2]. On 11 February 2025 the court granted Thomson Reuters partial summary judgment on direct copyright infringement and rejected Ross's fair use defence^[2]. Summary judgment at that stage means the court resolved the question on the record before it rather than sending it to a full trial.

Why it matters

The case turned on a boundary that many AI developers have had to think about: whether text written by a publisher about other people's documents — summaries, annotations, organising schemes — can be used as training material without a licence. Thomson Reuters had refused a licence request before the tool was built^[2], and the court's rejection of the fair use defence gave the company a win on the infringement question on the record before it.

For the company itself, the ruling protected the value of the editorial layer around Westlaw^[2] rather than only the raw text of judicial opinions. For readers following AI and copyright, it is one of the clearer data points on how courts in the United States may treat commercial training on a publisher's own editorial content.