On Tuesday, the American District Court Judge Delaware Stephanos Bibas issued a partial judgment in favor of Thomson Reuters in a matter of violation of copyrights against Ross IntelligenceLegal AI startup. Intricate in 2020, this is one of the first cases that will cope with the legality of AI tools and the way they are trained, often using data protected by copyright scraped elsewhere without a license or permission.
Similar lawsuits against OpenAI, Microsoft and other AI giants are currently going through the courts and can be reduced to similar questions about whether AI tools can apply for the defense of “allowed use” of using materials protected by copyright.
In the statement given The Verge Thomson Reuters, Jeff McCoy, the company said:
We are glad that the court issued a summary judgment in our favor and came to the conclusion that the editorial content of Westlaw created and maintained by our lawyer editors is protected by copyright and cannot be used without our consent. Copying our content was not “allowed”.
However, as the judge noted, this matter concerned the “non -generation” artificial intelligence, not the AI generative tool, such as LLM. Ross closed in 2021, Calling the lawsuit “false” But saying that he was not able to raise sufficient funds to continue when he consumed a legal battle.
As he reports earlier WiredToday Judge Bibas he wrote in his decision“None of the possible defense of Ross maintains water” against accusations of a violation of copyright, and eventually rejected Ross’s defense, relying largely on the factor in the employ of author’s materials by Ross, influenced the market for the values of the original work, building a direct competitor.
Thomson Reuters sued Rossa from the Westlaw search engine. Westlaw indexes a lot of materials that are not subject to copyright (like legal decisions), but also intertwine them with their own content. For example, Westlaw Headnotes – which are summaries of law points written by human editors – are a characteristic feature that is to make a very high-priced subscription to Westlaw attractive to lawyers.
By building a legal research search engine, Ross has transformed annotations and headnotes “into numerical data with relationships between legal words, to persuade his artificial intelligence,” wrote the Bibas. The ruling describes how Thomson Reuters rejected his attempt to licensing the content of Westlaw, Ross turned to another company, Legease and acquired 25,000 collective notes of questions and answers written by lawyers, using the head of Westlaw, whom he used for training data.
CEO of Ross Andrew Arruda claimed Westlaw data is “added noise” and that its tool “aims to recognize and separate the answer directly from the law by means of machine learning.” However, after “comparison, as similar to each of the 2830 questions of mass notes, heading and judicial opinions are one by one”, the judge said that evidence of actual copying was “so obvious that no reasonable jury could not be found differently”.