Artificial intelligence can help law firms work faster, but it also creates questions about who owns the content behind a tool. The first step in understanding the AI copyright risks for law firms is to know what software is doing and what your firm is asking it to do. Thomson Reuters v. ROSS gives firms a reason to take those questions seriously.
The Third Circuit Court of Appeals upheld a copyright decision against ROSS Intelligence on 29 September 2026. The ruling concerns a legal research product trained on proprietary editorial content from Westlaw. For firms, it offers practical lessons about choosing providers and handling content.
What happened in Thomson Reuters v. ROSS?
Thomson Reuters owns Westlaw, which includes court opinions alongside publisher-written material. That added material includes headnotes, which are short summaries of legal points in a decision. The dispute focused on those summaries.
ROSS wanted to build an AI legal research service that would compete with Westlaw. Thomson Reuters declined ROSS’s request to license its content. ROSS then obtained training memos from LegalEase, whose writers used Westlaw headnotes to prepare legal questions.
ROSS used the memos to train its search tool. Thomson Reuters sued, arguing that this process copied protected parts of its work.
What did the latest ruling decide?
A federal judge in Delaware issued the February 2025 ruling finding copyright infringement involving 2,243 headnotes and rejecting ROSS’s fair use defense. The September 2026 appeals decision upheld that result.
The distinction between the underlying law and a publisher’s writing is central here. Judicial opinions aren’t protected by copyright, but a publisher’s original summaries can be. Writing your own explanation of a decision is different from copying someone else’s protected explanation.
Why didn’t ROSS’s copying qualify as fair use?
Fair use permits certain uses of copyrighted work without permission. Courts consider the purpose of the use and the nature of the original work. They also consider how much was taken and the effect on the copyright owner’s market.
The appeals court saw ROSS’s service as a commercial competitor performing a similar legal research function. Training software didn’t give the copying a sufficiently different purpose. The court also found potential harm to Westlaw’s business and the market for licensing headnotes as AI training data.
That doesn’t mean every commercial use fails. Fair use depends on the facts, and a new technology doesn’t settle the issue by itself.
Does this decision ban AI use in law firms?
ROSS’s product wasn’t generative AI that writes fresh responses. It returned relevant passages from existing court opinions. The case therefore doesn’t decide whether every generative AI training practice is lawful.
The ruling also doesn’t say a law firm infringes copyright simply by using an AI service. A firm’s own conduct still needs a separate review. Uploading protected content or publishing copied output can raise different questions from running a legal search.
How should firms evaluate AI providers?
A practical response to AI copyright risks for law firms is to ask providers direct questions about their content. Find out whether the tool relies on licensed sources or public-domain material. Ask how it handles claims that its training data includes protected work.
Review the contract too. Does it offer protection if someone brings a copyright claim arising from your permitted use? Look closely at any limits on that protection and any restrictions on what you can upload.
These steps don’t guarantee that a tool is free from legal risk. They can help your firm compare providers and avoid relying on a broad assurance that all AI training is fair use. Keep the provider’s answers with your contract for future review.
Also ask what happens if the provider loses access to an important content source. A firm that depends on one research system should know whether it can export its work and how service changes will be communicated. This is a business continuity question as well as a legal one.
What if a firm builds its own AI tool?
The same questions deserve attention if your firm builds an internal AI system. Start by recording the source of each set of documents and confirming what uses are allowed. Don’t assume that a research subscription includes permission to reuse a publisher’s summaries for training.
For example, a firm might want a chatbot that searches its own research library. That library could contain attorney-written notes alongside copied excerpts from paid publications. Review those categories before uploading the collection and get permission where needed. Using the tool only inside the firm doesn’t automatically resolve copyright questions.
How does this affect law firm website content?
The ROSS case involved training a research tool, rather than writing a practice page. Even so, a useful lesson for marketing teams is to pay attention to where source content comes from. An AI tool shouldn’t become a shortcut for copying another firm’s work.
For example, feeding a competitor’s article into a chatbot and requesting a lightly reworded version can create problems. Changing words doesn’t necessarily remove protected expression. A stronger process starts with authoritative sources and develops an explanation suited to your firm’s audience.
Have a person check the finished draft against its sources. Confirm the legal statements and review any suspiciously familiar passages before publication. Keep notes showing where information came from and how the final content was developed.
What should an internal AI policy cover?
A written policy can turn these questions into daily habits. It should identify approved tools and explain what staff may upload. It should also assign responsibility for reviewing work before anyone uses it with a client or publishes it.
Copyright is one concern. Confidentiality needs its own review because client information can be exposed through a tool’s data practices. Before uploading it, check whether the provider stores submissions or uses them for further training. Determine whether informed client consent is required under the rules that apply to your practice.
The American Bar Association’s Formal Opinion 512 discusses duties involving competence and confidentiality when lawyers use generative AI. It also addresses supervision and reviewing output for errors. The guidance is based on the ABA Model Rules, so firms should check their jurisdiction’s requirements.
How can Civille support your firm’s marketing?
Thoughtful technology use starts with a clear process and people who understand it. Your firm can gain efficiency while setting limits on the content it uses and reviews. Those decisions should connect to a website that gives prospective clients clear, useful information.
As you evaluate AI copyright risks for law firms, Civille can help strengthen your website and content strategy. We provide law firm website design and SEO services, with marketing tools built for legal practices. Contact Civille on (920) 903-5168 to discuss your firm’s goals and next steps.
