The NYT v OpenAI copyright lawsuit just had its most dramatic week yet. On Thursday, unredacted filings in the three-year-old case — brought by The New York Times against OpenAI and Microsoft — became public, and honestly, they read less like a legal document and more like a confession. Inside the companies' own walls, executives were describing AI training on news content as theft, while their lawyers were publicly arguing the exact opposite.
01.What the unredacted filings actually say
The headline quote comes from Brent Hecht, Microsoft's director of Applied Science. In a January 2023 internal memo, he described the mass copying of web content for AI training as "an astonishing theft of unprecedented proportions" — and then went further, calling it "the largest theft of labor in human history." Not a critic. Not a competing publisher. Microsoft's own senior researcher.
On the OpenAI side, Nick Turley, who led ChatGPT, wrote in internal communications that publishers faced an "existential threat" from chatbots, which he described as "largely substitutive" — meaning they replace the original article rather than complement it. That word, substitutive, matters enormously in copyright law, because substitution is precisely what the fair use defence says isn't happening.
02.The numbers are staggering
The filings put actual counts on the copying for the first time. OpenAI's mid-training datasets alone contained more than 91,692 copies of works published by the NYT, Daily News and the Center for Investigative Reporting. A dataset built from Common Crawl held more than 2 million documents from nytimes.com. And a Microsoft initiative called Project Mango — through which Microsoft allegedly passed training data to OpenAI — assembled a dataset containing copies of at least 160,903 unique works from the news publishers.
Satya Nadella, Microsoft's CEO, testified in a deposition that "anything that is paywalled should be licensed by anyone who wants to use it... for grounding or training," and said that had he known OpenAI scraped paywalled content, he would have invoked Microsoft's right to require OpenAI to retrain its models. That's the CEO of OpenAI's biggest partner, on the record.
03.The 93% traffic drop
Here's the number that should scare every publisher — including India's newsrooms. According to the filing, Microsoft's own data showed its Copilot "answer engine" caused click-through rates for The New York Times' domain to drop by as much as 93% compared to traditional Bing search. An internal Microsoft presentation from January 2024 described the decline as a "doom loop" that would "hurt the performance of our models and the entire web at the same time."
Think about that loop for a second. AI answers are trained on publisher content. Those answers stop readers from visiting publishers. Publishers die. And eventually the AI has nothing new to learn from. Even Microsoft saw it coming.
04.Paywall hacks and stripped copyright notices
The filings also describe how the content was allegedly gathered. When OpenAI researcher Nick Ryder told president Greg Brockman about a "hack to get around nytimes paywall," Brockman's reply was: "ah nice." There are also allegations that copyright notices were deliberately removed from training data, because researchers "wouldn't want model outputting" those notices to users — a claim that cuts to the heart of the Digital Millennium Copyright Act, since removing copyright management information is a violation in itself.
If some of this sounds familiar, it's because we covered the Copilot DMCA ruling earlier — that case turned on the same question of stripped attribution. The courts are now hearing variations of this argument from multiple directions.
However the ruling lands, the money keeps flowing — investors are valuing OpenAI at $1.2 trillion.
05.OpenAI's defence isn't dead yet
To be fair, OpenAI hasn't conceded anything. The company argues that training on publicly accessible web pages qualifies as fair use, that it held an implied licence because publishers didn't initially block crawlers via robots.txt, and that the process is transformative — like search engines indexing the web. The US Department of Justice has also filed a Statement of Interest backing the fair use position on competitiveness grounds.
And the publishers' side keeps growing: a coalition represented by Platkin LLP now includes more than 550 publications, from the Tampa Bay Times to local weeklies. With discovery closing, the judge is expected to rule on the core fair use questions soon.
06.Why this matters for India
You might think this is an American courtroom drama with nothing to do with us. It isn't. Indian publishers — from big media houses to solo bloggers and YouTube educators — face the exact same dynamic: AI models trained on their work, answering readers directly, and cutting the traffic that pays the bills. Whatever precedent this case sets will shape licensing norms globally, including how Indian content gets valued (or taken) by frontier labs.
And if you're building with AI — freelancing, running an agency, shipping a product — this case decides whether "the model scraped it" becomes a real defence or a liability. Knowing how to use AI tools responsibly, with proper sourcing and cost discipline, is quickly becoming a professional skill, not a nice-to-have.
Key Takeaways
- Microsoft's own research director called AI scraping "the largest theft of labor in human history" in a 2023 internal memo
- OpenAI datasets held 91,692+ copies of plaintiffs' works; one Common Crawl dataset had 2M+ nytimes.com documents
- Copilot cut NYT click-through by up to 93% — Microsoft internally called it a "doom loop"
- Nadella testified paywalled content should be licensed and said he'd have forced OpenAI to retrain
- 550+ publications have joined the suits; OpenAI maintains its fair use defence, backed by a DOJ statement
Frequently Asked Questions
What do the newly unredacted NYT v OpenAI filings show?
They show a Microsoft exec privately calling AI scraping “the largest theft of labor in human history,” and reveal OpenAI's training data held 91,692+ copies of NYT, Daily News and CIR works.
How much NYT content was in OpenAI's datasets?
One dataset had 2 million+ documents from nytimes.com alone.
What did Microsoft's own data show about Copilot's impact on publishers?
Microsoft's data showed Copilot cut NYT click-through rates by up to 93% versus traditional Bing search. Over 550 publications have now joined consolidated suits, while OpenAI still insists scraping is fair use.
What This Means for You
If you're a student or professional using AI daily, this case will decide the rules of the tool in your hands — who gets paid for the content it learned from, and how much of the web survives to train the next generation of models. For Indian creators and small publishers, it's a signal to start protecting your work: robots.txt rules, clear terms, and licensing conversations instead of silent scraping. And for anyone building a career around AI, the winners will be those who use these tools with judgment — sourcing carefully, spending smartly. Our GPT-6 Astra — Zero to Hero Guide covers exactly that side of the equation, including a full cost-optimization chapter that keeps your AI usage sustainable as prices and rules evolve.