A recent ruling from a US appeals court has stirred the ongoing debate over copyright law and its relationship with artificial intelligence. The Third Circuit court upheld a decision stating that Ross Intelligence, an AI legal research company, violated copyright law by using materials from Thomson Reuters’ Westlaw database to train its legal search engine. This ruling sheds light on the complexities of the fair use doctrine under US copyright law.
The fair use doctrine, established under the US Copyright Act of 1976, allows copyrighted material to be used for various purposes, including criticism, journalism, and research. However, determining whether a use qualifies as fair use hinges on several factors: how transformative the use is, how much of the work is utilized, and whether its reproduction impacts the market value of the original. In this case, the court found that Ross effectively copy-pasted Thomson Reuters’ “headnotes” – brief editorial summaries of legal issues – directly into its tool, which the court ruled was not transformative.
Ross shut down in 2021, shortly after the lawsuit was filed, citing financial strains caused by the litigation. Ross shut down in 2021, shortly after the lawsuit was filed, citing financial strains caused by the litigation. Dozens of similar lawsuits have been filed by authors, comedians, the music industry, influential newspapers, and even Encyclopedia Britannica, all accusing AI companies of unlawfully using their intellectual property to train AI models.
How influential the Third Circuit ruling will be is debatable – the AI industry itself called it a win, with tech industry group Chamber of Progress senior director of AI Adam Eisgrau arguing the decision “implicitly confirms that highly transformative gen AI training to produce a hugely multi-purpose model with substantial public benefit is likely fair use.” The court’s decision specifically addressed the unauthorized use of headnotes to create a competing legal research tool, rather than issuing a blanket ban on generative AI training. “Ross took the headnotes to make it easier to develop a competing legal research tool,” noted a Delaware federal court judge in the original ruling last year. This focused approach may leave room for different interpretations regarding other forms of generative AI training.
The vestiges of Ross have vowed to take the lawsuit to the Supreme Court, though whether that would settle the broader legal battle is a tossup. Copyright Lately’s Aaron Moss argued the case isn’t the best vehicle for a sweeping ruling on generative AI: it reached the Third Circuit on interlocutory review, damages remain unresolved, Ross no longer exists, and its AI wasn’t generative.
Beyond this ruling, rights holders have other avenues to pursue. Moss noted that there are “cleaner cases for deciding the future of generative AI training” already scheduled for hearings. For example, a hearing for Concord Music Group’s lawsuit against Anthropic – which accuses the company of unlawfully using thousands of copyrighted song lyrics to train its Claude AI model – is scheduled for later this month.
