Can AI Take Your Work Without Paying?

Artificial Intelligence (AI) constantly consumes data to improve its capabilities. When that data includes someone else's copyrighted material, does using it constitute copyright infringement?

One court just said it can be. Thomson Reuters, which runs the Westlaw legal research platform, sued a legal tech startup called Ross Intelligence for copyright infringement. Ross Intelligence had used Westlaw’s legal headnotes—short summaries of court rulings—to train its AI. The judge ruled that Ross’s use of the copyrighted material was not fair use and violated copyright law.

Let’s look at what the ruling means and how it may impact businesses and creators.

Ross’s Use of Westlaw for AI Training

A distinctive aspect of AI is that it needs to be “trained,” which is accomplished by feeding it large amounts of data to help it get better at recognizing patterns, retrieving information, and (in the case of generative AI) generating responses. The more data it ingests, the better the AI gets.

Ross Intelligence was building an AI-powered legal research tool. To do this, it needed quality legal summaries to train its model. So, without a Westlaw license, Ross commissioned a contractor to create training materials based on Westlaw’s headnotes. Though researchers did not copy headnotes verbatim, their materials apparently did track their structure and phrasing.

The Court’s Ruling: Training Is Infringement (at Least Here)

Thomson Reuters sued Ross Intelligence, contending that Ross was wrongfully using its copyrighted material without permission to build its AI legal research tool. Ross’s response was essentially that AI training isn’t the same as copying—which is what copyright law prohibits.

Initially, the judge set the case for trial. But after reviewing the evidence more closely, the judge reversed course and made an unequivocal ruling: “Ross infringed 2,243 headnotes.”

The Court’s Reasoning: Training Is Not Fair Use (at Least Here)

The court found that the headnotes were creative enough to merit copyright protection and that Ross’s use of them constituted infringement. Ross had converted the headnotes into numerical data to train its AI models. The court reasoned that the conversion of text into data for AI training is an intermediate step in copying, as opposed to mere “learning,” as Ross had argued.

Ross also argued that its use was protected as “fair use,” which allows copyrighted material to be used without permission on a limited basis for things like commentary, criticism, and teaching. Ross argued that its use was “transformative,” meaning that it changed the material and gave it a purpose different from the original intended purpose. But the court rejected this argument, too, finding that Ross was building its tool for the same purpose as Westlaw: commercial legal research.

Takeaways for Creators and Businesses

This ruling is important for creative professionals—from musicians and filmmakers to architects and software developers—whose work may be targeted by AI. It offers some ammunition in creators’ battles with corporations that perennially seek to profit from their work without permission or compensation.

For businesses, the decision provides an early sign that AI development may require that training data be licensed rather than assuming that copyrighted works can be used freely. While many legal issues remain unresolved, companies that rely on AI tools and solutions should consider whether their providers have secured the right permissions.

Some further takeaways:

• Copyright still matters: The court didn’t give AI a free pass. Creatives may take this as a step in the right direction about their ability to control how their work is used.
• Commercial intent matters: The commercial purpose of the products weighed against fair use. If AI turns someone else’s work into a competing product, that could be a problem.
• Legal risks loom: Companies developing and using AI face legal uncertainties. They should review their data sourcing practices and consider licensing alternatives.

It’s important to note that this is just one case. The debate over AI and copyright is fair from over. The court emphasized that Ross’s AI was not generative AI. Businesses  and creates should stay informed as cases involving different technologies and different industries wend their way through the courts.