OpenAI-NYT paywall case impacts fair use for publishers

What Executives Actually Said

OpenAI’s head of ChatGPT, Nick Turley, stated publishers face an existential threat from OpenAI’s products, which he described as largely substitutive. He noted these products would become more and more substitutive as they improved. OpenAI co-founder Greg Brockman claimed their models are excellent at news and very good at any news task.

Microsoft CEO Satya Nadella testified that chatbots substitute website information, removing the need to visit the original source entirely. Internal Microsoft analysis backed this up with numbers: an 83 to 93% drop in click through rates for The New York Times and Daily News domains via Copilot compared to traditional Bing Search. The filings also mention executives discussing gazillions of dollars at stake in AI commercialization.

Why the Emails Hurt the Fair Use Defense

Legal experts believe the unsealed documents are devastating for OpenAI and Microsoft’s fair use defense. The fourth fair use factor concerns the impact on the market for the original work, and that’s exactly where these emails create problems.

David Buttle, founder of Spur, a publisher AI licensing coalition, emphasized that a real licensing market already exists. He noted that OpenAI, Microsoft, Amazon, Google, Meta, Perplexity, and Mistral have all signed publisher licensing deals for similar content and uses. Buttle argues the defendants’ own conduct establishes a market price, which makes the fair use argument harder to sustain when the companies have already paid for comparable access elsewhere.

Danielle Coffey, CEO of News/Media Alliance, added that Nadella’s testimony strengthens the Times’ argument that these products are genuinely substitutive rather than transformative.

The Paywall Evidence Changes the Calculus

Evidence of OpenAI bypassing paywalls moves the case beyond fair use gray areas entirely, according to Alan Chapell, privacy expert and founder of Chapell and Associates. He described this as clearly stolen territory, since paywall circumvention effectively admits the AI system substitutes the publisher’s product rather than merely learning from it.

Such actions almost always defeat a fair use defense on their own. Any deliberate paywall circumvention could also violate the Digital Millennium Copyright Act, the 1998 federal law that governs unauthorized access to protected digital content.

What This Sets Up for Other Publishers

This case could set a precedent for AI training lawsuits well beyond OpenAI and Microsoft. An anonymous publisher executive stated that if the Times wins, there will be lots of suits against companies like Google, Anthropic, and Meta.

• Coffey believes publishers will be ready to either sue or collectively license their content once precedent is set
• Buttle views large scraping cases, including the Reddit-Perplexity lawsuit, as a broader test for whether a real AI licensing market exists
• A strong ruling for rights holders would raise the cost and risk of unlicensed scraping, pushing developers toward structured licensing deals instead

What Comes Next

More unredacted documents are expected to emerge as the case proceeds. Legal teams for both sides will challenge what becomes public, and Judge Sidney Stein will ultimately decide those disputes, a process that could take months.

Both The New York Times and OpenAI/Microsoft are seeking summary judgment, arguing the discovery evidence is clear enough for the judge to resolve key issues without a jury trial. Coffey anticipates a mountain of evidence still to come before that happens.

Hashlytics Take

The paywall evidence is the part that should worry OpenAI’s legal team most, not the executive quotes. Admitting your product substitutes journalism is a business risk. Circumventing a paywall to get the content in the first place is closer to a confession, since it removes the “we’re just learning from publicly available information” argument entirely. If the paywall bypass holds up as described, this case stops being about whether AI training counts as fair use and starts being about whether OpenAI took content it was explicitly denied access to. That’s a much easier case for a publisher to win.