A judge has ordered that a jury decide a dispute over the use of artificial intelligence learning from published content, in what will be one of the first copyright trials over the technology’s unauthorized use of data.
The case was brought by news and information company Thomson Reuters against Ross Intelligence, which shut down it legal research platform in 2021 after the UK-based outlet sued, citing the costs of litigation.
At issue is whether Ross used Thomson Reuters’ legal-research platform Westlaw to train its own AI-based platform.
“Facts can be messy even when parties wish they were not,” U.S. Circuit Judge Stephanos Bibas wrote in his order that the case go to trial. He wrote that “many of the critical faces in this case remain genuinely disputed,” which is why he denied both parties’ request that he issue a summary judgment rather than bringing in a jury to hear the case.
Bibas wrote that “summary judgment is proper only if factual messes have been tidied. Courts cannot clean them up.”
A host of other tech companies, including Facebook and Instagram parent Meta Platforms, Microsoft-backed OpenAI and Stable Diffusion creator Stability AI have also been hit with lawsuits from copyright owners over the use of their work to train generative AI software, Reuters noted.
No date was set for the trial.
Thomson Reuters brought the suit in 2020, accusing Ross of copying Westlaw’s “headnotes,” which summarize points of law in court opinions, as part of its effort to build a competitor for the popular legal database. The suit claimed Ross misused “thousands of the headnotes to train its AI-based legal search engine,” Reuters said.
Ross claims its use of the Westlaw material was “fair use,” a concept that could become a pivotal question for legal disputes in cases involving generative AI training, the report said.
The judge wrote that a jury will decide the fair use question, along with other issues related to how far the Westlaw copyright extends over the material it publishes.
He did grant Thomson Reuters argument that Ross copied material from Westlaw, but said a jury will decide if the Ross materials meet a standard of “substantial similiarity” to the original material.
Bibas also said he could not decide whether ruling for Ross or for Thomson Reuters would best serve the public interest.
“Here, we run into a hotly debated question,” Bibas said. “Is it in the public benefit to allow AI to be trained with copyrighted material?”