What the unredacted filings actually say
The New York Times filed its copyright suit against OpenAI and Microsoft three years ago. A motion filed on behalf of the Times, flagged by digital media advocate Jason Kint, has now been unsealed and unredacted. The documents show executives at both companies privately described AI scraping as theft and their own products as a threat to the publishers whose material trained them. TechCrunch reported the unsealing on 17 September 2026.
The motion quotes Microsoft director of applied science Brent Hecht calling the AI buildout "an astonishing theft of unprecedented proportions" and "the largest theft of labor in human history." Those phrases were written internally rather than under oath. The unredacted text converts them into exhibits. Attorneys for the Times use the quotes to argue the industry built its foundation on an unfathomably large pile of appropriated intellectual property.
The filings also cover conduct, not only sentiment. They document paywall circumvention, including OpenAI President Greg Brockman's "ah nice" reply on learning of a hack to bypass the New York Times paywall. Microsoft CEO Satya Nadella testified that paywalled content should be licensed for AI training. The admissions span both companies' leadership and both the training and deployment sides of the products at issue.
The executives' own words: Hecht, Turley, Brockman, Nadella
Hecht's language is the sharpest on record. As Microsoft's director of applied science, he described copying the internet to train models as the largest theft of labor in human history, per India Today. The phrasing matters because it comes from the applied-science side of Microsoft, the person closest to what large-scale collection actually involves. An internal characterization of theft is harder to dismiss than an external critic's.
OpenAI's Nick Turley called chatbots "largely substitutive" and an "existential threat" to publishers. That is a product executive characterizing his own category as replacement rather than complement. Nadella went further under oath: he said he would have required retraining had he known OpenAI scraped paywalled material. A sitting CEO stating a retraining trigger on the record narrows the space for a defense built on innocent collection practices, per BigGo Finance.
Brockman's "ah nice" response to the paywall bypass is small in isolation. In a copyright suit it is not: it timestamps awareness at the president's level and pairs an admission of threat with an apparent welcome of circumvention. The filings do not state whether these statements were known to either company's legal teams before redaction, or surfaced only through this unsealing. That gap is itself a fact worth holding onto.
The 93% click-through collapse and what comes next in the case
Internal data in the filings showed click-through rates for news sites dropped as much as 93% in Copilot results. That number is the damages theory in miniature: if the product substitutes for the source at that rate, licensing stops being an abstract question. Turley's "largely substitutive" characterization and the 93% figure point the same direction from different rooms — one from product strategy, one from measurement.
The case has run three years, and the unsealing indicates discovery is producing material the defendants did not want public. The documents reveal the theft admission and the threat AI products pose to publications. The record here is silent on settlement posture and trial scheduling; the filings speak to liability framing, not outcome. Readers should treat the quotes as evidence about intent, not as a verdict.
Two signals matter from here. First, Nadella's testimony that paywalled content should be licensed gives publishers a CEO-level endorsement of the exact licensing model they are suing to obtain. Second, the internal Copilot data supplies a quantified substitution metric that can anchor damages arguments. The open question is whether courts treat executive admissions about theft as evidence of intent, or as sentiment irrelevant to a fair use analysis.
Sources
- techcrunch.com — Microsoft Exec Called AI Scraping 'the Largest Theft of Labor in Human History'…
- biggo.com — Microsoft and OpenAI Executives Admitted AI Poses ... (2026-09-17)
- indiatoday.in — AI companies copying internet to train models biggest theft of labour in history… (2026-09-18)
- futurism.com — Microsoft Director Privately Admitted AI Was the "Largest Theft of ... (2026-09-17)
- New Horizon daily digest, 2026-09-18 (topic provenance)