Unsealed Filings Show OpenAI and Microsoft Feared an AI 'Doom Loop'

The News

Newly unsealed filings in the New York Times' copyright lawsuit against OpenAI and Microsoft, reported by The Verge on 18 September, show both companies privately grasped the damage their chatbots were doing to the open web. The 92-page document quotes senior figures conceding that scraping the internet to train large language models amounted to, in the words of Microsoft's Director of Applied Science Brent Hecht, the "largest theft of labour in human history".

Hecht also wrote that Microsoft's defence made "a complete mockery of the idea of 'fair use'". Microsoft has since distanced itself: spokesperson Alex Haurek said the remarks reflected "one employee's individual perspective" and were "not a legal analysis". A separate filing from Jordan Usdan, a general manager at Microsoft AI, cast Hecht's role as deliberately adversarial and academic.

The document goes further. An internal Microsoft memo warned that its content strategy had triggered a "doom loop" threatening both model quality and the wider web. OpenAI cofounder Greg Brockman was quoted eyeing the "gazillions" of dollars on offer, while chief executive Satya Nadella conceded that chatbots had largely replaced search.

Why It Matters

The heart of the dispute is what publishers call "Google Zero" - the point at which AI answers strip away the click-throughs that fund journalism. OpenAI's own experts reportedly estimated that search referrals to sites such as the Times may be down as much as 60 percent, a decline they linked to AI summaries including Google's AI Overviews.

That admission matters because it undercuts the fair-use argument propping up the entire generative-AI business model. When rival products launched, the industry insisted training was transformative and harmless to rights holders. Internal notes describing ChatGPT as "the modern newsstand", and Head of ChatGPT Nick Turley reportedly saying there is "no good reason to click" a source, suggest the firms knew better. The last comparable flashpoint was the 2023 wave of publisher lawsuits; this filing is the first to expose the private candour behind the public confidence.

Indian Angle

For India, the case lands on fertile legal ground. ANI Media is already pursuing OpenAI in the Delhi High Court in the first significant copyright challenge of its kind in the country, and the Federation of Indian Publishers and the Digital News Publishers Association - whose members include the largest media houses - have sought to join the fight. The unsealed US admissions hand their counsel fresh ammunition.

Indian law may prove less forgiving than America's. The Copyright Act, 1957 permits only narrow "fair dealing" under Section 52, a far tighter standard than US fair use, and OpenAI has argued that Indian courts lack jurisdiction because its servers sit abroad. Evidence that the companies internally doubted their own defence could weaken that position considerably.

There is a commercial dimension too. Indian newsrooms and regional-language publishers lean heavily on search and aggregator traffic; a 60 percent collapse in referrals would hit thinly funded vernacular outlets hardest. MeitY, which is drafting the Digital India Act, may find the documents useful as it weighs whether AI firms should license Indian content rather than harvest it freely.

FAQ

What is the New York Times case about?

The Times sued OpenAI and Microsoft in December 2023, alleging their models were trained on its articles without permission and can reproduce them word for word. The newly unsealed filings form part of that continuing litigation in a US federal court.

What is "Google Zero"?

It is shorthand for a web in which chatbot answers satisfy users so completely that they never click through to the original source. That severs the referral traffic and advertising revenue publishers rely on, and internal estimates in the filing put the referral decline for some sites at as much as 60 percent.

Does this affect the Indian ANI case?

Not directly, but the admissions could be cited persuasively. ANI Media's suit against OpenAI in the Delhi High Court raises similar questions about unlicensed training data, and Indian publishers and industry bodies have been watching the US disclosures closely as they weigh their own claims and possible interventions.

Where can I read the original coverage?

The Verge published the full breakdown of the 92-page filing, including the internal quotes from Hecht, Nadella and Brockman, on 18 September 2026. Its report reproduces images of the relevant passages and links to the underlying court documents.

This story was reported by The Verge. Read the full original coverage at The Verge.