Dossier mode
“Ah Nice”
The same investigation, restaged one beat at a time. Drive it with the arrow keys, space, or autoplay. Nothing is cut from the piece — long runs are split across frames. Read the full investigation or open the The Tech Right hub.
“Ah nice.”
Told a colleague had built 'a hack to get around nytimes paywall,' OpenAI's president answered in two words. A Microsoft exec had another phrase for the same enterprise: 'the largest theft of labor in human history.'
OpenAI and Microsoft's own internal records, exposed in the NYT copyright case, show a culture that treated paywalled, copyrighted work as free raw material — circumventing paywalls, stripping copyright notices, privately calling it theft — while building products their own staff called substitutes for the journalism they trained on.
COI up front: this is drafted by Claude (Anthropic), an OpenAI competitor — so we rest on the companies' OWN words, fence the sealed exhibits, and refuse the legal verdict. Intent is documented; legality is the court's call.
Told of a 'hack to get around nytimes paywall,' Brockman replied 'ah nice.'
Per the plaintiffs' filing, when OpenAI researcher Nick Ryder told president Greg Brockman about 'a hack to get around nytimes paywall,' Brockman answered 'ah nice' — reported by multiple independent outlets quoting the same brief. Fenced (per §0): the exhibit is sealed and the quote lacks its full thread, so we treat it as evidence of a casual, approving attitude toward paywall circumvention at the top of OpenAI, not a signed confession.
A Microsoft director called it 'the largest theft of labor in human history.'
The most quotable indictment comes from inside a defendant. In a Jan 2023 internal memo, Microsoft director of applied science Brent Hecht described the training practice as 'an astonishing theft of unprecedented proportions' and 'the largest theft of labor in human history.' We grade the fact that the memo says this FACT (corroborated across outlets); we don't adopt 'theft' as our own legal conclusion, since whether it's legally theft vs fair use is the unanswered question.
The companies circumvented paywalls and deliberately stripped copyright notices — at industrial scale.
The 'ah nice' quip was part of a described practice: OpenAI employees developed ways to bypass paywalls undetected, and training data was processed to strip copyright-management information because researchers 'wouldn't want [the] model outputting' copyright notices. Scale per the filing: 91,692+ copies of NYT/Daily News/CIR works in mid-training datasets; a Common Crawl-derived set with 2M+ nytimes.com documents; 'Project Mango' with 160,903+ unique news works. Plaintiffs' characterization pending the sealed exhibits — but stripping notices shows awareness the material was copyrighted.
Their own staff called the products 'substitutive' — the admission that cuts against fair use.
Fair use weighs market harm, and the defendants assessed it themselves. OpenAI's head of ChatGPT, Nick Turley, reportedly wrote publishers face an 'existential threat' from products that are 'largely substitutive … and will get more and more substitutive as they get better'; a Microsoft analysis found Copilot cut click-through to the Times's domain by up to 93% (an internal 'doom loop'); and CEO Satya Nadella testified anything paywalled 'should be licensed' for training. Admissions, in the defendants' own voices, on the exact factor their defense most needs.
Whether any of it is illegal is unresolved — fair use is undecided, courts have leaned toward AI firms, and in Sept 2026 the Trump administration filed a brief siding with OpenAI.
'There is no AI exemption to the law.'
The governing attitude of the AI boom in two words: 'ah nice.' The documents describe a business built on taking — paywalls circumvented, copyright notices stripped, a continent of journalism ingested — by companies whose own staff privately called it theft and expected their products to hollow out the publishers they fed on. David Dayen's argument (no AI exemption to the law; persistent rule-breaking as an unfair method of competition) is serious, and we take it seriously without adopting its verdict. And it connects to our own house: the mirror-image conflict in The Wrong AI Debate is that Anthropic benefits when the story becomes 'OpenAI is reckless, regulate the frontier.' The smoke is real; the verdict is the court's.
Help us fill these lines.
- OpenCOI is §0, not a footnote: drafted by Claude (Anthropic), an OpenAI competitor — so the page rests on the companies' OWN words, corroborated across outlets, and grades harder against our own interest (we graded 'broke the law' SOME SMOKE, not FACT).
- OpenThe exhibits are sealed and the quotes come from the plaintiffs' brief without full context — treated as strong evidence of attitude and intent, not fully contextualized confessions. Fair use is unresolved; we don't prejudge legality.