Could a seemingly obscure copyright ruling about building codes become the lynchpin in the high-stakes legal war between legal tech titans? This week, the 3rd U.S. Circuit Court of Appeals ordered Thomson Reuters and the defunct ROSS Intelligence to file supplemental briefs focusing on the implications of its American Society for Testing & Materials v. UpCodes, Inc. decision. The result? Predictably, both sides are now screaming from the rooftops that the opinion vindicates their own narrative. It’s a classic legal skirmish played out in the digital age, with each party trying to weaponize precedent.
The core of the dispute, as always, boils down to how AI can learn from existing data without infringing copyright. ROSS, the startup that famously sought to build an AI legal research platform, sees the UpCodes ruling as a lifeline. Thomson Reuters, the established giant defending its proprietary data, views it as just another piece of evidence demonstrating ROSS’s missteps. The differing interpretations are as stark as the courtroom itself.
ROSS: The Fair Use Renaissance is Here
For ROSS, the UpCodes decision is nothing short of a “controlling” precedent that necessitates reversing the lower court’s ruling against it. Their argument hinges on the idea that AI innovators can indeed copy available legal materials to construct new research tools, and that fair use permits this. It’s a bold play, framing their ambition as a forward-thinking technological leap rather than outright infringement.
They draw direct parallels: West (a Thomson Reuters product) is the industry leader selling subscriptions to access copyrighted materials, much like ASTM. ROSS, the startup that used this content without a license to power its AI, mirrors UpCodes, the defendant in the other case. Both employed freemium models and offered AI-driven tools. It’s a narrative designed to fit neatly into the UpCodes framework.
ROSS leans heavily on the first and fourth factors of the fair use analysis. They argue the district court’s initial assessment of ROSS’s “subjective purpose” was flawed. The UpCodes court, they contend, clarified that the inquiry must be objective – looking at the function of the use, not the user’s internal thoughts. Furthermore, ROSS asserts its use was “at least as transformative as what UpCodes did,” fundamentally changing the medium and using the data to train an AI model, not just to present it directly.
On market harm, ROSS claims the district court erred by defining the relevant market too broadly. It wasn’t the market for “legal research platforms” generally, but specifically for headnotes. And crucially, West has offered no concrete evidence of lost subscribers or revenue directly attributable to ROSS. The UpCodes decision, ROSS argues, demands a granular examination of the record, not mere inferences of harm from competition.
“In 2020, West should have responded to ROSS with innovation,” the brief concludes. “Instead, West chose litigation. Copyright rewards the former, not the latter.”
This closing salvo is pure rhetorical firepower, aiming to paint Thomson Reuters as an obstructionist rather than an innovator.
Thomson Reuters: The Line in the Sand Remains Unmoved
Thomson Reuters, predictably, sees things through a diametrically opposed lens. For them, the UpCodes decision changes nothing substantial. Instead, it reinforces the fundamental distinction in fair use: between commercial uses that substitute for the original and uses that serve a truly different purpose without supplanting the original market. And in their view, ROSS’s actions fall squarely into the former category.
They argue that while UpCodes found UpCodes’ use transformative because its purpose (showing what the law is) differed from ASTM’s (publishing standards), ROSS’s purpose was to create a competitor. This isn’t a difference in purpose; it’s a direct challenge to Westlaw’s core business. Thomson Reuters emphasizes that UpCodes protects uses that are transformative and non-substitutive, neither of which, they claim, describes ROSS’s activities.
The Central Dispute Crystallized
Here’s the thing: both ROSS and Thomson Reuters are experts at legal rhetoric. Their briefs are masterclasses in selectively highlighting favorable points and downplaying inconvenient facts. But the UpCodes ruling has inadvertently forced the core issue into starker relief. It’s not just about whether AI can learn from copyrighted data; it’s about the nature of that learning and whether it creates a derivative product that directly competes with, or harms the market for, the original work.
My unique insight here? The legal system is still grappling with how to apply old copyright doctrines to nascent AI technologies. The UpCodes case, and now this appeal, highlight the tension between encouraging innovation and protecting existing intellectual property. The courts are trying to find a balance, but as we’ve seen time and again in tech, the law often plays catch-up.
Thomson Reuters’ argument that UpCodes only applies to uses that don’t “supplant the original work’s markets” is a critical distinction. ROSS’s claim of transformative use—training an AI—runs headfirst into TR’s assertion that this training process was designed to build a direct, substitutive competitor. The devil, as always, is in the details of how this transformative use actually manifests in the end product and its market impact.
The Unanswered Question: Intent vs. Impact
Ultimately, this appeal boils down to whether the court prioritizes the intent behind AI development (training a model) or the impact of the resulting product (potentially displacing existing services). ROSS’s lawyers are skilled enough to argue that their intent was purely for transformative learning, and any market impact was incidental or indirect. Thomson Reuters, of course, is arguing the opposite: that the intent was to build a commercial rival, and the market impact was inherent in that design.
It’s a complex dance, and this latest round of briefing shows that the UpCodes decision, while seemingly about building codes, has opened a Pandora’s Box for AI-driven legal tech. We’re not just talking about legal research anymore; we’re discussing the fundamental rules of how AI learns and competes in the digital economy.
The stakes are enormous, impacting not just these two companies but the entire future of AI development in sensitive fields like law, medicine, and finance. Whether UpCodes truly compels a reversal for ROSS, or merely reinforces Thomson Reuters’ position, remains to be seen. But one thing is clear: the legal battles over AI and copyright are far from over.
Is the UpCodes Ruling a Win for AI Innovation?
While ROSS touts the UpCodes decision as a victory for AI innovation, Thomson Reuters argues it merely clarifies existing fair use principles, distinguishing between genuinely transformative uses and those that directly substitute for the original work’s market. The interpretation hinges on whether ROSS’s AI training process is seen as creating a fundamentally different purpose or a direct competitor.
Why Does This Matter for Legal Tech?
This litigation has massive implications for the legal tech industry. It defines the boundaries of permissible data usage for AI model training, impacting how new AI-powered legal tools can be developed and deployed. A ruling favoring ROSS could spur further innovation, while a win for Thomson Reuters might necessitate more licensing agreements and potentially slow down the development of competitive AI legal research platforms.
