Newly unsealed court documents in a major copyright lawsuit reveal that insiders at Microsoft and OpenAI privately worried that generative artificial intelligence could undermine the publishers and creators whose work helped train their systems. Internal Microsoft material described mass AI scraping in extraordinarily stark terms and warned of a potential commercial “doom loop”: AI systems absorb publishers’ content, provide users with answers that reduce visits to original websites, weaken publishers’ revenues and ability to produce new material, and ultimately degrade the supply of quality information upon which AI models themselves depend. The documents, cited by news organizations suing Microsoft and OpenAI, also contain internal concerns about AI products substituting for journalism and disrupting creative employment. Microsoft and OpenAI maintain that training AI models on copyrighted material can qualify as fair use, while Microsoft has emphasized that the statements attributed to one employee reflected his individual perspective rather than the company’s position. The litigation could help determine where American copyright law draws the boundary between technological innovation and the property rights of those who create the material that makes advanced AI possible.
Key Takeaways
- Internal Microsoft documents warned that generative AI could create a “doom loop” by reducing traffic and revenue for publishers, eventually damaging the content ecosystem on which AI systems themselves depend.
- Microsoft data cited in court filings reportedly found that its Copilot answer engine reduced click-through rates to one major news publisher by as much as 93 percent compared with traditional Bing search, strengthening publishers’ argument that AI-generated answers can function as substitutes for original content.
- The dispute raises a fundamental property-rights question for the AI economy: whether companies may ingest copyrighted material at massive scale under fair use or must license and compensate creators when their commercial systems are built upon—and potentially compete against—the creators’ work.
In-Depth
The artificial-intelligence revolution is colliding with an old American principle: people who create valuable property generally expect to control how others commercially exploit it. Newly unsealed court filings in copyright litigation against Microsoft and OpenAI suggest that people inside the companies understood this conflict long before it became a public battle.
The documents quote Microsoft Director of Applied Science Brent Hecht describing mass scraping for AI training in exceptionally severe terms. More consequentially, Microsoft researchers reportedly recognized a structural danger to their own business model. If AI answers replace visits to websites, publishers lose advertising, subscription, and referral revenue. With less revenue, publishers produce less original material. AI companies then have less high-quality human-created information available for training and grounding future systems—the “doom loop” described internally.
That concern is no longer theoretical. Court filings cite Microsoft data indicating Copilot reduced click-through rates to one publisher’s domain by as much as 93 percent compared with conventional Bing searches. OpenAI executives also reportedly discussed the possibility that increasingly capable chatbots could become substitutes for publishers.
Microsoft and OpenAI dispute the publishers’ legal theory. AI companies have maintained that training models on copyrighted works can constitute fair use, an argument with significant implications for American technological competitiveness. Microsoft has also said Hecht’s comments represented an individual perspective rather than corporate policy.
The courts must therefore confront a difficult balance. America has a legitimate interest in leading artificial-intelligence development, but innovation does not automatically extinguish intellectual-property rights. If AI companies depend commercially on journalism, books, photography, and other human creations, determining where fair use ends and compensable appropriation begins may become one of the defining property-rights battles of the digital age.
Sources
- https://www.theepochtimes.com/us/tech-giant-feared-its-ai-tools-started-a-commercial-doom-loop-new-documents-reveal-6094128
- https://www.washingtonpost.com/business/2026/09/17/microsoft-exec-called-ai-largest-theft-labor-history-court-records-show/
- https://techcrunch.com/2026/09/17/microsoft-exec-called-ai-scraping-the-largest-theft-of-labor-in-human-history-new-unredacted-filings-reveal/
- https://arstechnica.com/tech-policy/2026/09/microsoft-exec-called-ai-scraping-the-largest-theft-of-labor-in-human-history/