What the first major US AI copyright ruling means for IP law

Copyright claims against AI companies have received potential boosts.

Last week, US federal judge Summary judgment was given In a case that technology conglomerate Thomson Reuters raised against legal tech company Loss Intelligence. The judge found that Ross would use Ross’ content to train AI legal research platforms that are infringed on Reuters’ intellectual property.

The results may have more impact than this 39 Copyright-related AI lawsuits He is currently passing through US courts. That said, it is not necessarily a slam dunk for plaintiffs who claim that AI companies have infringed their IP rights.

All the headlines

Ross was accused of using headnotes (a summary of legal decisions) by Westrow, a Reuters legal research service, to train AI. Ross sold AI as a tool to analyze documents and perform query-based searches across court filings.

Ross argued that the use of copyrighted headnotes is legally defensible as it is transformative. In other words, we have reused our headnotes to serve significantly different features or markets. In his summary judgment, Stefanos Vivas, the judge who is leading describing the case, did not find the argument particularly convincing.

Ross, in his opinion, is repackaging Westrow’s principal in a way that directly replicates Westrow’s legal research services, Vivas said. The startup platform did not add any new meanings, purposes or commentary.

In his decision, Vivas also cited Ross’ commercial motivation as the reason the startup defense missed the mark. Ross tried to benefit from products that competed directly with Westlaw without the critical “recontextualization” of IP-protected Westlaw material.

Shubha Ghosh, a professor at Syracuse University who studies IP methods, calls Thomson Reuters a “strong victory.”

“The trial will continue (but) Thomson Reuters has been awarded a summary judgment, a victory at this stage of the lawsuit,” Gauche said. “The judge also confirmed that Ross is not entitled to summary judgment on his defense, such as fair use or mergers. As a result, the case continues to be tried with a strong victory for Thomson Reuters.”

The application is small

Already there is at least one set of plaintiffs in another AI copyright case I asked the court to consider Vivas’ decision.. However, it is not yet clear whether precedents will shake up other judges.

Bibas’ opinion emphasized that it distinguishes between “generated AI” and the AI ​​that Ross used, but it didn’t generate any content, but spouted out the already written judicial opinion.

Generic AI is at the heart of copyright lawsuits against businesses Openai and Mid Journeyis frequently trained on a large amount of content from public sources on the web. Supplying many examples, generative AI can generate speeches, text, images, videos, music, and more.

Most companies developing generator AI claim it Doctrine of fair use Protects the practice of scraping and using it for training without compensation or crediting the owner of the data. They claim that they have the right to use the content they have been published in training, and that their models are essentially outputting transformational work.

However, not all copyright holders agree. Some also point to a phenomenon known as RefluxGenerate AI creates content that is very similar to the trained work.

Randy McCarthy, a US patent attorney for the law firm Hall Estil, said Vivas focused on “the market impact of the original job.” However, he also warned that Vivas’ opinion was relatively narrow and could have been overturned by appeal.

“One thing is clear, at least in this case. You can simply use copyrighted material as training data, and AI cannot be used in itself fairly,” McCarthy tells TechCrunch He spoke. “(But) one fight in a massive war. We need to look at more developments before we can extract the laws related to using copyrighted materials as AI training data. ”

Another attorney, TechCrunch, spoke with Mark Lezama, a litigation partner at Knobbe Martens, focusing on patent disputes. He views that judge reasoning can be extended in many ways into generated AI.

“The court refused to defend fairly as a matter of law as Ross used Chief (Thomson Reuters) to develop a competing legal research system,” he said. “The court suggested that this may be different from the situation involving generative AI, but because generative AI competes with news sites using copyrighted articles, it is important to train the generative AI. It’s easy to see news sites that claim that copying articles is the same for users’ attention.”

In other words, publishers and copyright holders are ruling it with AI companies, and there are few reasons to be optimistic after the decision. A little.