The U.S. District Court in Delaware has delivered a significant ruling in a case involving artificial intelligence trained on copyrighted Thomson Reuters’ data. Judge Stephanos Bibas ruled that Ross Intelligence had gone too far. This judgement, while not final, is a milestone in the legal world regarding the protection of content in machine learning.
Copyright dispute: Ross vs. Thomson Reuters
In May 2020, Thomson Reuters, the owner of the Westlaw research platform, filed a lawsuit against the start-up Ross Intelligence. Westlaw is a popular online research service for lawyers available in over 60 countries and containing more than 40,000 data entries, including case law, statutes, codes and articles.
The platform is designed for efficient and accurate online legal research and analysis that can be done from anywhere. It has even partnered with OpenAI in the field of advanced legal technologies to improve its product and use artificial intelligence.
In the lawsuit, Thomson Reuters alleged that Ross illegally used Westlaw’s data, including two key types of information: headnotes, which are brief summaries of key case aspects, and the Key Number System classification tool, which is used to organise the topics of legal questions.
Initially, Ross Intelligence attempted to obtain a licence to use these materials. Ultimately, the company used intermediaries who provided data from Westlaw without Thomson Reuters’ consent. The court found that 2,243 headnotes had been copied from Westlaw, which was the basis for deciding the case.
Copyright and fair use
In order to decide the case, two fundamental questions had to be answered:
- Did Ross Intelligence infringe copyright?
- Was Ross Intelligence acting within the bounds of fair use?
Ross’ lawyers argued that their system did not create content, but merely facilitated searches. Users did not have access to the originals, only to results that were not linked to the source material.
In February 2025, Judge Bibas issued a partial summary judgment finding that Ross Intelligence had infringed Thomson Reuters’ copyrights by using copyrighted material without the copyright owner’s permission. The judge found that Ross’ actions were commercial in nature and aimed to create a product that would compete with Westlaw, which could have significantly damaged Thomson Reuters’ strong market position.
As a result, the judge found that Ross’ actions went well beyond the bounds of fair use. The premise of fair use is based on the fact that in certain situations it is possible to legally use copyrighted works without the owner’s consent, for example, to create parodies or for non-commercial research.
In addition, the court emphasised that copying the copyrighted material was not necessary to achieve the goal of training the AI. Instead, Ross was attempting to reproduce the creative work of Westlaw’s editors, which was found to be unjustified. Questions regarding the use of the Key Number System were left for further consideration.
High risk of unlicensed use of data
This landmark ruling could be highly significant for future cases involving the use of copyrighted content by AI. The ruling shows that the commercial use of data without the owner’s consent is a copyright infringement, even in the context of training artificial intelligence.
Although Ross Intelligence closed down in 2021, largely due to the high cost of pursuing the case, the ruling is significant for small technology companies, as a cautionary tale for those considering using data without proper licences. This case is different from others involving generative AI. Ross’ intention was not the creation of content, but only the improvement of search for content.
Publishers’ damages claims
Other AI-related cases are still on US judges’ desks. The New York Times, for example, is suing OpenAI and Microsoft, alleging that the companies used its articles to train AI models without permission.
Similar lawsuits have been filed by newspaper owners such as Alden Global Capital and music companies such as Concord Music Group, seeking compensation for the use of their content to train AI.
The decision in Thomson Reuters v Ross Intelligence sets a precedent in the US and may have an impact on other disputes in this area. It highlights the risks for companies using unlicensed data and raises questions about the limits of fair use in an era of advanced technological development.
Are you concerned that your database application is infringing copyright?
Not sure if your company is allowed to use the data it collects?
Contact us to avoid unwanted disputes.
