New Evidence in NYT Lawsuit: OpenAI Internally Acknowledged AI Posed 'Existential Threat' to Publishing Industry
nashnova research
The New York Times filed new evidence showing OpenAI internally acknowledged its AI products pose an "existential threat" to publishers — yet still copied millions of copyrighted articles for training. This multibillion-dollar lawsuit has entered its most consequential phase.
What did OpenAI say internally?
The head of ChatGPT wrote internally that publishers face an "existential threat" from AI products because they are "largely substitutive, and will become increasingly so."
This means → OpenAI was not caught off guard. It knowingly chose to train on copyrighted content while acknowledging the product could replace the very outlets it was copying.
Co-founder Greg Brockman's internal remarks from around 2017 were blunter: he was "deeply motivated by the insane amount of wealth" commercializing OpenAI's technology could generate.
How was the paywall bypassed?
New filings reveal an employee told Brockman he had "found a way to get around the New York Times paywall." Brockman replied: "Oh, nice."
In plain terms = the Times built a paywall to charge for its journalism; OpenAI used technical means to skip it — and senior leadership knew and tacitly approved.
Times lawyers argue the methods were "transgressive and deceptive," raising questions not just about copyright but about the legality of how the content was obtained.
Why is Microsoft caught up in this?
Microsoft's director of applied sciences stated internally that an OpenAI win would "make a complete mockery of the concept of 'fair use'" — effectively Microsoft's own staff questioning the fair-use defense.
CEO Satya Nadella testified that had he known OpenAI scraped content behind paywalls, he would have invoked his contractual rights to require OpenAI to retrain its models.
The most damaging number: users of Microsoft's AI answer tool click through to the underlying source content 83% to 93% less often than traditional search users. This means → AI delivers the answer directly, so users never visit the original article — the most concrete evidence of substitution on record.
Can OpenAI's "fair use" defense still hold?
OpenAI has consistently argued that training models on publicly available content qualifies as "fair use" — a legal doctrine allowing limited use of copyrighted material without permission under certain conditions.
The company previously called its models a "transformation," not a "copy," of the original content.
But the new evidence builds a logical chain: internal acknowledgment of substitution + knowing copying + paywall bypass + Microsoft data showing users no longer click through. This reflects a fair-use defense under factual siege from multiple directions.
What does this case mean for the entire industry?
This is the most prominent of dozens of copyright suits filed against AI companies. Its outcome will directly define the compliance boundaries for training data across the sector.
In plain terms = if the Times wins, every AI company will have to re-examine where its training data comes from; if OpenAI wins, the scope of fair use expands dramatically.
A reference point: the Financial Times signed a licensing and partnership agreement with OpenAI in 2024 — some publishers have opted to negotiate first, not litigate.
市场有风险，内容仅供研究参考，不构成投资建议。