AI COUNSEL NEWS + PODCAST
AI COUNSEL Podcast
ORAL ARGUMENT - Thomson Reuters v. ROSS Intelligence (The Appeal)
0:00
-52:33

ORAL ARGUMENT - Thomson Reuters v. ROSS Intelligence (The Appeal)

53 minutes of Colloquy on AI Copyright + Fair Use Elements

Haven’t listened yet , other than the awkward opener when Judge calls “Thomson ROOTERS” and he or someone says “ROSS.” [They said it in ALL CAPS.]

Here a summary from Grok below

As have been reporting BIG BIG IMPLICATIONS FOR FAIR USE TRAINING WHAT IS LEGAL CONTENT WHAT IS NOT FOR AI FOR LAW THIS IS A FUTURE LODEBEARING CASE

THE “FIRST” LAWBOTS!

******************************MAY LISTEN AND DO COMMENTARY LATE THINGS BIZZY THIS WEEK HOOBOY WITH HTTP://LIQUIDLAW.AI THE BEST WEEK TO LAUNCH PRIVILEGED COMMS AND AI INFRA FOR LAW! PARDON MY LESSER ABILITY HERE ON AI COUNSEL NEWS + PODCAST

Ruling expected in (months?)

Thomson Reuters Enterprise Centre GmbH and West Publishing Corp. v. Ross Intelligence Inc. (D. Del. No. 1:20-cv-00613-SB) is a significant U.S. copyright case involving the use of Westlaw materials to train a competing AI legal research tool.

Background

Thomson Reuters owns Westlaw, a major legal research platform that includes judicial opinions organized via the proprietary Key Number System and original editorial content such as headnotes. Headnotes are concise summaries of key legal points from cases, created through editorial selection, judgment, and expression.

Ross Intelligence developed a competing AI-powered legal research search engine that allowed users to query in natural language and retrieve relevant case law. Ross sought a license for Westlaw content but was refused because it was a direct competitor. Instead, Ross contracted with a third party (LegalEase) to create approximately 25,000 “Bulk Memos” — compilations of legal questions and answers. These memos were prepared by lawyers who relied heavily on Westlaw headnotes. Ross then used the Bulk Memos to train its AI model.

bpb-us-e2.wpmucdn.com

Thomson Reuters sued Ross in May 2020 for copyright infringement (primarily of the headnotes), among other claims.

Key Rulings

2023 Summary Judgment Opinion

Judge Stephanos Bibas (U.S. Circuit Judge sitting by designation) largely denied both parties’ summary judgment motions. He found genuine factual disputes on most issues, including whether headnotes were protectable, whether copying occurred, substantial similarity, and fair use.

The case was set to proceed to trial.

February 11, 2025 Revised Opinion

On renewed motions, Judge Bibas revised his earlier ruling and granted partial summary judgment in favor of Thomson Reuters:

  • Direct Copyright Infringement: Ross infringed Thomson Reuters’ copyrights in 2,243 headnotes. The court found:

    • Headnotes are original and copyrightable (they involve creative editorial expression, not just facts or the underlying judicial opinions).

    • Actual copying occurred (Ross had access via the Bulk Memos; the memos closely tracked headnote expression more than the raw opinions).

    • Substantial similarity existed for these specific headnotes.

    • Copyright validity (e.g., possible expiration or registration timing issues) for these headnotes remains a jury question.

      bpb-us-e2.wpmucdn.com

  • Fair Use Defense: Rejected
    The court held that Ross’s use was not fair use as a matter of law. Fair use analysis (17 U.S.C. § 107):

    Factor

    Ruling

    Reasoning


    1. Purpose & Character

    Strongly favors Thomson Reuters

    Commercial use to build a competing product. Not transformative — Ross used the headnotes for the same purpose (legal research assistance) rather than a new purpose or to access unprotected ideas.


    2. Nature of the Work

    Slightly favors Ross

    Headnotes are mostly factual/editorial summaries with limited creativity (less than highly original works like novels).


    3. Amount & Substantiality

    Favors Ross (somewhat)

    Intermediate copying for training; headnotes were not output to end users. Not necessarily the “heart” of the works in the final product.


    4. Market Effect

    Strongly favors Thomson Reuters

    Direct competitive harm. Ross created a market substitute and undermined potential licensing markets for AI training data.

    Factors 1 and 4 (the most important) outweighed the others. Overall balance: No fair use.

  • Other Defenses Rejected: Innocent infringement, copyright misuse, merger doctrine, and scènes à faire.

Current Status (as of June 2026)

Ross appealed the February 2025 rulings via interlocutory appeal (§ 1292(b)).

The Third Circuit granted permission to appeal in June 2025 (Case No. 25-2153).

Oral argument was held on June 11, 2026.

The district court proceedings are stayed pending the appeal.

No final resolution has been issued by the Third Circuit yet.

Significance

This is one of the earliest district court decisions directly addressing whether using copyrighted material (specifically expressive headnotes) to train a commercial AI tool that competes with the original product qualifies as fair use.

The ruling emphasizes that commercial, non-transformative uses that create direct market substitutes are unlikely to be fair use, even if the copying is intermediate and the protected material is not shown to end users.

It has been widely discussed in the context of broader AI training copyright disputes.

This is a summary of the district court’s February 2025 opinion (the key operative ruling).

The case remains on appeal, so the final outcome could change.

Discussion about this episode

User's avatar

Ready for more?