OpenAI and Microsoft knew they were starting a ‘doom loop’ for the web

Terrence O’Brien’s Verge report walks newly unsealed documents from the New York Times’ case against OpenAI and Microsoft. Internal language is stark: Microsoft’s Brent Hecht called ChatGPT and Copilot’s data harvesting the “largest theft of labor in human history” and said the fair-use defense made a “complete mockery” of the idea; Microsoft now distances itself, calling Hecht’s views individual and academic, not company policy.

An internal Microsoft document warned the AI content strategy had started a “doom loop” that would hurt model performance and the entire web by threatening the economic foundations of the content supply chain. Other highlights: Greg Brockman on “gazillions” of commercial upside; an OpenAI representative “unaware” of efforts to detect or remove paywalled content despite Nadella saying paywalled material should be licensed; employees noting GPT-4 memorized heavily and would be “insanely good at regurgitation,” with examples of long verbatim outputs from the Times and other outlets; Nick Turley on users having “no good reason to click” source links; OpenAI experts linking publisher referral drops—speculating search referrals may be down as much as 60%—to AI summaries.

We’re for AI assistants that cite, license, and leave the web alive—and these filings show the companies saw the wreckage coming.

“Doom loop” is not activist rhetoric; it is their phrase for a product that substitutes for the labor that feeds it. Distancing Hecht after the fact does not erase the supply-chain memo or the regurgitation admissions. Knowing paywalls were porous while talking licensing principles is the tell.

Useful AI can summarize the world without zeroing publishers into oblivion. That requires paid access, provenance, and click-worthy citations—not “modern newsstand” framing that ends the click. Google Zero was predicted in their own docs; treating it as an accident insults the record.

I’m watching how judges weigh these quotes, whether licensing deals scale beyond trophy publishers, and whether product defaults change to send traffic back instead of finishing the doom loop.

The Verge published on September 18, 2026, covering a 92-page filing. Microsoft spokesperson Alex Haurek said Satya Nadella’s testimony on how people find information should not be confused with Microsoft’s legal copyright positions.