Mother Jones reports that documents cited in its copyright lawsuit against OpenAI and Microsoft offer an inside view of how executives discussed AI training data, publishers’ rights and the economic effects of chatbots. The lawsuit argues that the companies built valuable AI products using copyrighted material without permission or compensation. An internal Microsoft document reportedly warned that large models were “hoovering up” human work and could create a “doom loop” by weakening the suppliers that produce online content. Company documents and research cited in the filing say AI search reduced clicks to news articles by 90 percent compared with traditional search, while referral traffic from Google Search and Discover fell at least 30 percent after AI overviews were introduced. The article says OpenAI’s data efforts specifically targeted high-quality content, that news was especially prevalent in one training dataset, and that the company aimed to “crush freshness” in real-world news. The filing also alleges that datasets included paywalled articles despite commercial-use restrictions and that copyright notices, author information and terms of use were stripped from collected pages. According to the article, OpenAI created a filter after the lawsuits were filed to prevent models from reproducing plaintiffs’ content; the brief argues that the filter was intended to limit evidence of copying rather than stop infringement generally. An analysis of 20 million ChatGPT responses found more than 400,000 identical 25-word sequences with Mother Jones articles, although the article notes that some overlaps could reflect shared quotations rather than direct regurgitation. It also recounts Greg Brockman praising ChatGPT’s ability to complete New York Times text and responding positively to a paywall-bypass technique. The plaintiffs argue that free use of content gives each company an incentive to continue even though the industry as a whole could benefit from compensation, creating a collective “doom loop” that may degrade the internet and its content supply. These are allegations and arguments in ongoing litigation, not final court findings. Mother Jones concludes that decisions about content rights and AI’s broader direction should not be left solely to AI executives.
