In This Article
Newly unsealed filings in The New York Times’ copyright case reveal internal Microsoft and OpenAI discussions about AI training, news scraping and the impact of generative AI on publishers. The statements sharpen the debate over whether training on copyrighted journalism is fair use, but they are not a court ruling or necessarily company-wide positions.
Key takeaways
- Microsoft and OpenAI employees raised concerns about the effects of AI training and products on publishers and the open web.
- The Times-led publishers argue that news content was used without permission and that AI answers can substitute for original reporting.
- Microsoft and OpenAI maintain that their use of content is permitted by copyright law; the dispute remains unresolved.
For enterprise data leaders, the filings also highlight why data provenance, governance and the long-term health of information sources matter—not just model performance.
Internal concerns become public
The filings include Microsoft applied science director Brent Hecht’s warning that people could view large-scale AI training as an unprecedented taking of creative labor. Microsoft has said his comments reflected an individual perspective, not the company’s legal analysis or official view. Internal Microsoft material also described a potential “doom loop”: AI products could reduce publishers’ traffic and revenue, weakening the sources that help supply future models with information.
OpenAI executives likewise discussed the effect on news organizations. ChatGPT head Nick Turley reportedly described the products as increasingly substitutive for publishers, while co-founder Greg Brockman remarked on the model’s ability to complete sentences from Times articles. The filing also describes internal discussion of getting around a news paywall to access content.
The legal questions remain open
The Times and other publishers allege that millions of articles were used to train AI systems without authorization and argue that AI-generated answers can compete with the journalism they reproduce or summarize. Microsoft and OpenAI counter that their use is transformative and protected as fair use. The companies’ internal conversations may be relevant evidence, but they do not by themselves establish infringement or settle the legal test.
The case, filed in 2023, includes publishers seeking damages. The plaintiffs have requested summary judgment, and a decision is not expected until 2027, according to reporting on the filings. Until the court rules, the boundaries of copyright law for AI training remain contested.
What data leaders should take from the case
The dispute concerns public web content and journalism, not every use of data in enterprise analytics. Still, it underscores the value of knowing where data comes from, what rights govern its use and how systems handle it over time. For organizations building reporting and AI capabilities, clear governance can reduce uncertainty and support more dependable decisions.
Spargent helps enterprises unify data and improve reporting with Microsoft Fabric and Power BI. In that work, documented data flows and well-governed analytics are practical foundations for scaling business intelligence—separate from, and no substitute for resolving, the copyright questions now before the court.
- “An astonishing theft of unprecedented proportions”: Court records show what Microsoft and OpenAI actually thought about AI training, Nieman Lab.
- Court Filing Shows Microsoft Exec Called AI Training the "Largest Labor Theft in Human History", gHacks.
- New Filing Shows OpenAI and Microsoft Debated AI Threats to the Web, The Mac Observer.
- Microsoft Exec Warned AI Web Scraping Makes A Mockery Of Fair Use, HotHardware.
- NYT alleges Microsoft, OpenAI knew using news content was theft, Tech Xplore.
