Unsealed filings reveal internal concerns over AI training on news articles
|
The Facts
- Newly unsealed court filings included internal communications from Microsoft and OpenAI employees.
- The New York Times and other publishers sued OpenAI and Microsoft over alleged copyright infringement.
- Publishers allege the companies used millions of news articles to train AI systems without authorization.
- Microsoft Director of Applied Science Brent Hecht described AI copying as an “astonishing theft” in internal communications.
- OpenAI and Microsoft argue that training AI on newspaper articles qualifies as fair use.
- The companies say their AI training transforms copyrighted material into new content.
- Publishers argue internal messages show the companies’ AI products could compete with journalism.
Context
What is the central legal dispute?
Publishers contend that OpenAI and Microsoft infringed copyrights by using news articles for AI training. The companies argue the use is protected by fair use because the training is transformative and does not replace the underlying journalism. Investing.com,mint,Straits Times
Why do the internal messages matter to the case?
The publishers say the communications describe AI products as substitutes for journalism, which they argue conflicts with the companies’ position that their products do not compete with news organizations. Investing.com,Yahoo News,mint
Who is Brent Hecht?
Brent Hecht is Microsoft’s director of applied science; court filings cited internal comments attributed to him about the use of published material in AI training. NDTV,Washington Post,Irish Independent
Where Left and Right agree, and where they split
Left and right largely agree on this one.
- Where Left and Right agree
- Both framings conclude the companies' own internal words undercut their fair-use defense, since an 'astonishing theft' admission contradicts claims of mere transformation.
- Where Left and Right differ in emphasis
- Both say the internal admission sinks the fair-use defense; the left stresses journalists going unpaid and uncredited, the right stresses courts must not excuse the contradiction as fair use.
- Why they won’t converge
- The divide is over values — how much weight copyright protection for creators' labor should carry against the innovation gains claimed from transformative AI training — a balance no shared timeline of facts resolves.
How left and right read it
Whether journalists get paid for the work behind these systems is the real stake, because millions of articles were taken to train tools that publishers say could compete with the newsrooms that produced them. That is why a Microsoft scientist's words calling the copying an "astonishing theft" undercut the transformation defense: the resulting product "conveys information to readers, without properly crediting -- or paying for -- the original work of journalists."
“Using this material, the AI giants have produced a behemoth that conveys information to readers, without properly crediting -- or paying for -- the original work of journalists.” — Washington Post
Fair use should mean real transformation, not wholesale copying — yet Microsoft's own Brent Hecht called the practice an "astonishing theft," undercutting the claim that these systems merely create new content rather than compete with the journalism they were trained on. Publishers whose articles were used without authorization deserve better than a transformation defense contradicted by the companies' own internal words. Courts should not let that contradiction be waved away as fair use.
“High-level staffers at a pair of top tech companies knew that scraping online news content to bolster their AI models could have devastating impacts on publishers - but that didn't stop them, according to a bombshell legal filing.” — New York Post
The receipts — all 68 sources
Wire services (3)
Independent coverage (50)
Facts first. Then every angle.
The day’s biggest stories in one short brief — the facts everyone agrees on, then the competing values behind the headlines. Free in your inbox.