Recent developments have provided a significant boost to copyright claims involving AI firms.
Last week, a U.S. federal judge issued a summary judgment in a legal dispute initiated by Thomson Reuters, a major technology conglomerate, against Ross Intelligence, a legal tech startup. The ruling determined that Ross’ incorporation of Reuters’ content to train its AI-driven legal research tool constituted a violation of Reuters’ intellectual property rights.
This ruling may have significant ramifications for the over 39 ongoing copyright lawsuits against AI companies in U.S. courts. However, this does not guarantee an easy path for plaintiffs who claim that their intellectual property rights have been breached by AI providers.
Understanding the Headnotes
Ross faced allegations that it utilized headnotes—concise summaries of legal judgments—from Westlaw, the legal research tool owned by Reuters, to train its artificial intelligence system. Ross presented its AI as an innovative solution for analyzing documents and conducting query-based searches within court filings.
In its defense, Ross contended that its employment of the copyrighted headnotes was legitimate because it transformed their use, adapting them for a different purpose or market. However, Judge Stephanos Bibas, presiding over the case, found this rationale unpersuasive.
According to Judge Bibas, Ross was merely reconfiguring Westlaw’s headnotes in a manner that closely mirrored Westlaw’s legal research service, failing to provide any additional meaning, purpose, or commentary that would support its claim of transformative use.
Furthermore, the judge noted Ross’ commercial interests as a critical factor undermining its defense. Ross intended to profit from a product that directly competed with Westlaw without substantially “recontextualizing” the intellectual property-protected Westlaw material.
Shubha Ghosh, a professor at Syracuse University specializing in intellectual property law, regarded this as a “strong victory” for Thomson Reuters.
“While a trial will still go ahead, Thomson Reuters has secured a summary judgment, marking a significant win at this stage of the litigation,” Ghosh noted. “The judge also asserted that Ross was not entitled to summary judgment on its defenses, such as fair use and merger. Thus, the case proceeds to trial, with a notable advantage for Thomson Reuters.”
Limited Applicability
At least one group of plaintiffs in a separate AI copyright case has requested that the court consider Bibas’ ruling. However, it remains uncertain whether this precedent will influence other judges.
Judge Bibas’ opinion distinctly differentiates between “generative AI” and the AI utilized by Ross, which simply reproduced previously written judicial opinions rather than creating new content.
Generative AI, which is frequently at the forefront of copyright litigation against firms like OpenAI and Midjourney, relies on vast datasets collected from public sources online. By analyzing extensive examples, generative AI can produce text, images, videos, music, and more.
Most companies involved in generative AI assert that fair use doctrines protect their practice of extracting data for training purposes without compensating or even acknowledging the original content creators. They believe their outputs qualify as transformative works capable of aiding artistic and creative progress.
Nonetheless, not all copyright holders are in agreement. Some highlight the issue of regurgitation, where generative AI produces outputs strikingly similar to the original works it trained on.
Randy McCarthy, a U.S. patent lawyer at Hall Estill, indicated that Judge Bibas’ emphasis on “the impact on the market for the original work” could prove pivotal for rights holders pursuing cases against generative AI developers. However, he cautioned that Bibas’ ruling is relatively narrow and could face reversal upon appeal.
“It is clear from this case that using copyrighted material as training data for AI does not automatically constitute fair use,” McCarthy remarked to TechCrunch. “[Yet, this is] one battle in a broader conflict, and further developments are necessary to clarify the legal standing regarding the use of copyrighted materials for AI training.”
TechCrunch also consulted with attorney Mark Lezama, a litigation partner at Knobbe Martens specializing in patent disputes, who believes that Judge Bibas’ ruling could have broader implications. He posits that the reasoning applied in this case could extend to various forms of generative AI.
“The court dismissed a fair-use defense as a matter of law partly due to Ross’s use of [Thomson Reuters] headnotes to establish a competing legal research system,” he explained. “While the court hinted that this might differ from scenarios involving generative AI, it’s plausible to envision a news organization arguing that utilizing its articles for training generative AI is similar, as the generative AI competes with the news outlet for audience engagement.”
In summary, publishers and copyright holders in conflict with AI firms have a slight cause for optimism following this decision—though it’s crucial to note the emphasis on slight.
Compiled by Techarena.au.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
