Takeaways from Thomson Reuters v. Ross Intelligence: AI Litigation Risks and Insurance Solutions

More than 25 lawsuits brought by copyright holders against AI companies are currently pending in federal court.^[1] These copyright holders generally allege that the use of their copyright-protected writings, music, voices, images and videos to train AI models violated their copyright rights. One of the earliest substantive decisions we have from these cases was recently issued by the court in Thomson Reuters v. Ross Intelligence. While every case might involve distinguishing facts, AI developers and users would benefit by: (1) analyzing this federal trial court’s decision that concluded an AI company’s use of copyrighted text to train its AI model violated copyright laws and was not protected by fair use;^[2] and (2) understanding the types of insurance policies that are available to protect against these lawsuits and the nuances associated with each type of policy.

Key Takeaways: 

1. The Federal District Court for the Eastern District of Pennsylvania (the Court) ruled that the defendant-AI developer’s use of copyrighted material as training data for its AI model was: (1) a copyright violation and (2) not protected by fair use. 
2. The Court justified its fair use decision by reasoning: (1) the AI developer used the copyrighted material to create a product that directly competed with the copyright holder; and (2) the AI developer’s use of the copyrighted material would undermine the copyright holder’s ability to license its data to other companies to train AI.
3. The Court’s decision raises concerns that fair use may not be a viable defense for AI developers in future lawsuits brought by copyright holders. 
4. AI companies should be aware of provisions within insurance polices that bar or limit coverage for copyright lawsuits. 

Background: Thomson Reuters v. Ross Intelligence

Thomson Reuters owns Westlaw, which is one of the largest legal research platforms in the world. Lawyers and other legal professionals pay Westlaw a subscription fee to access case law, state and federal laws, state and federal regulations, and law journals from Westlaw’s website. Westlaw also contains its own editorialized content called “headnotes,” which are short one-sentence summaries of key points of law or a court’s decision.

Ross Intelligence is a startup and competitor to Westlaw that used AI to train its legal search engine. Ross trained its AI by: (1) slightly rewriting Westlaw’s headnotes into the form of a question; and (2) inputting these revised questions into its AI model.^[3] Thomson Reuters discovered that Ross used its headnotes to train its AI model, and Thomson Reuters sued for copyright infringement. The Court concluded that Ross committed direct copyright infringement, finding it copied and materially appropriated Westlaw’s headnotes. Ross argued that, despite any alleged copyright infringement that may have occurred, its actions were protected by fair use. 

Fair Use Analysis: 

            When evaluating the fair-use defense, courts typically consider four factors: (1) the use’s purpose and character, including whether it is commercial or nonprofit; (2) the copyrighted work’s nature; (3) how much of the work was used and how substantial that part is, relative to the copyrighted work’s whole; and (4) how Ross’ use affected the copyrighted work’s value or potential market. In Thomson Reuters v. Ross Intelligence, the Court explained that it was going to give the most weight to the first and fourth factors.^[4]

            The first factor focuses on whether the use of copyrighted material by the alleged copier is commercial and transformative. The Court concluded the first factor favored Thomson Reuters because Ross’ use was commercial and non-transformative. There does not seem to have been much dispute on this point because Ross clearly sought to profit from the use of Westlaw’s headnotes as training data for its AI. Additionally, the Court concluded Ross’ use was non-transformative because Ross’ use did “not have a further purpose or different character from [Westlaw’s legal research platform.]”^[5] Specifically, the Court reasoned: “Ross was using Thomson Reuters’s headnotes as AI data to create a legal research tool to compete with Westlaw. . . . [And] when a user enters a legal question, Ross spits back relevant judicial opinions that have already been written.”

            The second factor measures “the degree of creativity inherent to the [copyright-protected] work.” More creative works get more protection. The Court noted this factor “rarely play[s] a significant role in the determination of a fair use dispute.” In any case, the Court concluded the second factor favored Ross and a fair use determination because: “the headnotes have creative elements but are far from the most creative works.”

            The third factor weighs “both the quantity of the materials used and their quality and importance.” In other words, to win this factor, “the alleged copier must not take the ‘heart’ of the work.” The Court concluded this factor weighed in favor of Ross because Ross’ output to the end user is a judicial opinion, not a Westlaw headnote, “so [Ross’s product] communicates little sense of the original.”

            For the fourth factor, courts analyze “how the [alleged infringer’s] use affected the copyrighted work’s value or potential market.” The Court noted the fourth factor is “undoubtedly the single most important element of fair use.” The Court analyzed the effect Ross’ use would have on: (1) the legal-research market; and (2) the market for the sale of training data to AI companies. The Court concluded that this factor should be weighed against fair use because: (1) Ross intended to create a market substitute for Westlaw; and (2) Ross’ use would damage Westlaw’s ability to license its data to AI companies for the training of AI models. The Court rejected Ross’ fair-use defense because the first and fourth factors were weighed in Thomson Reuters’ favor. 

Insurance Considerations for AI Developers and Users:

            Companies developing AI and those deploying it could face copyright claims. One step the companies can use to reduce their exposure to copyright and other intellectual property claims is purchase various types of insurance that cover the litigation costs and damages associated with copyright lawsuits. A number of policies have provided coverage for decades that might apply to copyright and intellectual property lawsuits, and the addition of AI to a lawsuit does not per se eliminate coverage. 

Commercial General Liability Insurance (CGL): CGL policies most commonly provide coverage to a business for personal injury and damage to property. But these standard form policies also provide “Personal and Advertising Injury” coverage that extends to liability from publication, including copyright infringement, misappropriation of another party’s ideas or advertising, defamation, and other tort claims.[6] Additionally, CGL policies commonly provide coverage for indemnity obligations, which could be valuable if the alleged liability is partially based on contracts the AI developer or AI user signed that contain indemnity obligations. The same policies also may contain exclusions that pertain to: (1) property damage arising out of the sale or licensing of computer software; (2) alleged infringement of certain intellectual property rights; and (3) professional services. None of these exclusions should defeat express coverage for copyright infringement, making these policies—commonly referenced as “litigation insurance”—valuable assets as the contours of AI-based liability evolves.

Technology Errors and Omissions Insurance: Technology errors and omissions polices cover the costs for damages due to errors, omissions, or negligent acts arising from the products or services provided by businesses. Many of these policies provide coverage for allegations of infringement. Sometimes these policies’ express coverage distinguishes between copyright, trade secrets and other information or works that may be claimed to be protected. As these policies are much less form-based than CGL policies, AI developers and deployers should look at what information they are using, how they are using it, and the terms of their policy.

Media Liability Insurance: These policies typically provide coverage for claims against the policyholder arising out of statements, information or data used in promotional material and advertising, but many times have broad coverage grants. Many such policies expressly provide coverage for copyright infringement, among other claims. These policies could provide coverage for claims against AI developers or deployers for copyright infringement. However, coverage may depend on the facts underlying the claims and what the policyholder allegedly did to give rise to a claim. 

Intellectual Property Liability Insurance: These policies are designed to address intellectual property infringement claims, such as copyright infringement. They also are often designed for technology companies that are developing and deploying software that might lead to infringement claims. The coverage for these policies can apply to defense costs, potential liability, and enforcement costs. These policies are less common than other forms of coverage and policyholders may need to consider the costs and scope of coverage. But these policies could be vitally important for small companies rolling out AI models that include allegedly copyrighted materials.

Cyber Insurance: Cyber insurance provides a range of coverage to help protect companies from, among other things, data breaches, cybersecurity issues, breach response, data restoration and privacy violations. Such coverage typically is triggered by a loss to the policyholder’s own data, software or other covered property, or a claim by a third party that it was harmed by a breach or other incident in which data was compromised. If an AI tool were to be a repository of copyrighted works and the AI data is compromised, several claims could arise that fall within coverage. However, without enhanced coverage for certain risks, cyber policies may exclude coverage for AI companies when liability is based solely on indemnity obligations. We also note that standard form cyber policies do not cover copyright claims, but cyber policies may be relevant for third-party claims against AI companies listed above. 

Insurance policies often contain limitations on coverage. As many policyholders have unfortunately experienced, broad coverage grants can be subject to sublimits and exclusionary language that narrows coverage—whether in definitions, conditions or express exclusions. As more businesses begin to use AI for an array of purposes, they will need to consider the risks of claims such as copyright infringement and other misappropriation claims. At the same time, they should consider the insurance protections they have in place and how they will respond to those risks.

While the Court’s decision in Thomson Reuters v. Ross Intelligence is not dispositive on whether courts will bar AI developers and AI commercial customers from using a fair-use defense in the future, the decision underscores the need for AI companies to address their intellectual property risks. AI companies should be prepared to seek coverage under their insurance policies if these risks come to fruition. 

[1] Kate Knibbs, Every AI Copyright Lawsuits in the US, Visualized, Wired (Mar. 9th, 2025, 10:30 AM), https://www.wired.com/story/ai-copyright-case-tracker/.

[2] Thomson Reuters v. Ross Intelligence, No. 1:20-cv-613-SB (E.D. Pa. Feb. 11, 2025).

[3] For example, Westlaw’s copyright headnote states: “Originality, for copyright purposes, means that the work was independently created and has some minimal degree of creativity.” Ross would use this headnote to train its AI model after slightly rewriting the Headnote to: “Does originality for copyright purposes mean that the work was independently created and has some minimal degree of creativity?” 

[4] Thomson Reuters v. Ross Intelligence, No. 1:20-cv-613-SB at 16 (E.D. Pa. Feb. 11, 2025) (citing Authors Guild v. Google, Inc., 804 F.3d 202, 220 (2d Cir. 2015)). 

[5] The Court explained that “Transformativeness is about the purpose of the use. If an original work and a secondary use share the same or highly similar purposes, and the second use is of a commercial nature, the first factor is likely to weigh against fair use[.]” 

[6] The term “standard form policies” refers to insurance policies that are used by many insurance companies with the policies’ terms remaining largely consistent among the different insurance providers. The consistent use of a standard form policy is due to the policy’s reputation as an industry model.