Photo via Washington Examiner

OpenAI and Microsoft leaders knew AI may cause ‘doom loop’ hurting entire web

Read the original at Washington Examiner ↗
Cite

Zimmermann, D. (2026, September 18). OpenAI and Microsoft leaders knew AI may cause ‘doom loop’ hurting entire web. Washington Examiner. https://www.washingtonexaminer.com/policy/technology/4732847/openai-microsoft-knew-ai-doom-loop-hurt-web/

Dated when The Aporia collected the article, usually within hours of publication. Check the original for its exact date and byline.

Washington Examiner · collected 2026-09-18 · by David Zimmermann

Quick Summary

Executives from OpenAI and Microsoft acknowledged that their AI technologies pose significant risks to the sustainability of news publications by diverting readership and income. In an internal document, Brent Hecht, Microsoft’s director of applied science, warned about a "doom loop" where AI models consume vast amounts of content for training, thereby hurting both the quality of these models and the web's overall health. The New York Times is suing OpenAI and Microsoft over copyright infringement related to using their copyrighted content without permission or payment, arguing that this practice undermines traditional business models reliant on subscriptions and advertising revenue.
Written by AI on 2026-09-18, from this article's own text only (the story summary below draws on all the coverage).
See how they differ ↓

AI analysis

Story summary

In a recent court filing unsealed in Manhattan federal court, executives from OpenAI and its major investor Microsoft admitted that large language models (LLMs) are undermining news publishers' business models through what they call a "doom loop." This loop involves LLMs using vast amounts of copyrighted content without permission, which then cannibalizes readership and revenue for the original publishers. An internal Microsoft document states this is an "unusual situation where an end-product threatens the economic foundations of its essential suppliers."

Executives like Brent Hecht, director of applied science at Microsoft, warned that AI models could harm both internet integrity and their own business by undermining news outlets’ sustainability. This admission comes as part of a long-running copyright infringement lawsuit filed by The New York Times against the companies over unauthorized use of copyrighted content for training LLMs like ChatGPT. The court documents reveal executives' private concerns that scraping millions of articles without consent amounts to "the largest theft of labor in human history." Despite these internal fears, Microsoft and OpenAI continue to defend their practices publicly by arguing they are protected under fair use laws since the AI transforms original content into new material.

Written for “AI Impact on Web Copyright” on 2026-11-20, grounded in this article and the 5 other(s) covering the same event.

How they differ

4 outlets covered “AI Impact on Web Copyright”.

What each outlet says

One AI summary per outlet of its latest article on this story, written from that article alone and checked against it. Open the article for the outlet's own words.

Washington ExaminerThis outlet18 Sep

Executives from OpenAI and Microsoft acknowledged that their AI technologies pose significant risks to the sustainability of news publications by diverting readership and income. In an internal document, Brent Hecht, Microsoft’s director of applied science, warned about a "doom loop" where AI models consume vast amounts of content for training, thereby hurting both the quality of these models and the web's overall…

OpenAI and Microsoft leaders knew AI may cause ‘doom loop’ hurting entire web (this article)
The Straits Times18 Sep

Karen Weise and Mike Isaac report on internal concerns at Microsoft and its partner OpenAI regarding the use of millions of copyrighted news articles to train AI systems. Documents reveal that Microsoft employees discussed fears that this practice might constitute a significant intellectual property theft, potentially harming journalism and leading to a "doom loop" where compromised data quality could undermine the…

Inside Microsoft and OpenAI, worry about damaging the publishing industry →2 articles from The Straits Times on this story
New York Post18 Sep

Microsoft and OpenAI employees are privately expressing concerns that the companies' practice of using millions of news articles to train AI systems constitutes a significant exploitation of intellectual property. Internal documents reveal fears that this could be viewed as an "astounding theft of unprecedented proportions." While one executive at OpenAI acknowledged the potential existential threat their products…

Microsoft, OpenAI workers privately worried their work is ‘largest theft of labor in human history’ →2 articles from New York Post on this story
404 Media17 Sep

Lawyers for The New York Times have unsealed an internal Microsoft document in a long-standing copyright lawsuit against OpenAI, revealing executives admit large language models (LLMs) are built on stolen content and represent a significant threat to human labor. These LLMs create what the document calls a "doom loop," where AI steals content from creators, cannibalizes their traffic, and destroys their business…

‘Doom Loop’: OpenAI and Microsoft Admits LLMs Are Destroying the Web and Built on Theft →

Where they agree and split

Framing

How the first headlines put it

Stated as fact4
The headlines (4)

‘Doom Loop’: OpenAI and Microsoft Admits LLMs Are Destroying the Web and Built on Theft 404 Media · stated as fact

OpenAI, Microsoft executives’ quotes on AI training threaten copyright defence, news outlets argue The Straits Times · stated as fact

OpenAI and Microsoft leaders knew AI may cause ‘doom loop’ hurting entire web Washington Examiner · stated as fact

Staffers at tech giants knew AI tools posed ‘existential threat’ to news publishers: court docs New York Post · stated as fact

Words to watch (1)

‘Doom Loop’: OpenAI and Microsoft Admits LLMs Are Destroying the Web and Built on Theft 404 Media

The full story: what happened, how it changed, and all 6 articles →

Compared with similar articles

This article reads leans left and hedges 20% of its claims. Each row says how that neighbour differs.
The Straits Times
Leans left Hedges 1 of 1 claims Outlet record 64/100
New York Post
Hedges 4 of 21 claims Outlet record 66/100
New York Post
Leans left Hedges 3 of 4 claims Outlet record 66/100
404 Media
Leans strongly left further left than this Hedges 1 of 20 claims Outlet record 95/100
Persuasion
Hedges 22 of 113 claims
“The articles discuss different aspects of AI risks and incidents; one focuses on rogue AI behavior accessing networks, while the other discusses internal warnings about AI's impact on the internet.”
The Straits Times
Leans left Hedges 1 of 14 claims Outlet record 64/100
“The articles discuss different aspects of AI development and concerns but do not describe the same specific incident or release.”
ABC News (AU)
Leans left Hedges 4 of 33 claims Outlet record 67/100
“The articles discuss different aspects of AI and copyright issues but do not describe the same specific incident or occurrence.”
404 Media
Leans left Hedges 6 of 45 claims Outlet record 95/100
“The articles discuss related issues with AI and its impact on spam and web content but describe different specific incidents or statements by executives.”
The Guardian
Hedges 1 of 34 claims Outlet record 64/100
“The articles discuss different aspects of AI and copyright issues without describing the same specific incident.”
CBS News
Hedges 5 of 18 claims Outlet record 67/100
“The articles discuss different aspects and warnings related to AI but describe distinct events with separate focuses.”

Story

AI Impact on Web Copyright
Technology · 6 article(s) covering the same event. Compare every outlet's account →
How this article was measured signals, leaning charts, publisher record, byline, topics and claims

Signals How these are calculated →

Claims extracted
15
sentences that state something as happening or true
Uncertain
20%
3 of 15 hedged ("reportedly", "may", "according to")
Leaning
Leans left
of the writing, not the subject · beta estimate
Publisher signals
67/100
Washington Examiner's corrections and hedging, across everything collected from it. Not a measure of accuracy.
Outlets on this story
6
Technology
Same framing elsewhere
1
articles that frame the event the way this one does
Analyzed 2026-09-18 · how these are computed

How this is being covered How these are calculated →

Article leaning vs. publisher reliability
Source leaning vs. consistency

Publisher

Washington Examiner · 5208 article(s) · 11 correction(s) detected
Running correction rate · 11 correction(s)

Who wrote this

David Zimmermann
161 article(s) here · 1 carrying a prediction
After serving two terms, Gov. Gavin Newsom (D-CA) will leave Sacramento in January 2027 to let Gov.-elect Xavier Becerra (D-CA) take his place.
Sen.-elect Scott Colom (D-MS) is set to become the fourth Senate Democrat who has an identical twin outside the upper chamber of Congress.
Treasury Secretary Scott Bessent directly blamed the Commerce Department for delaying the broader administration’s intent to place Nvidia customer Megaspeed on a special blacklist of companies that could pose a national security threat.
But now that she is out as chief of staff, it appears that the White House staff’s interactions with registered lobbyists will be cut down dramatically.
“We will continue to treat this with urgency until full mitigation, and we will share a full Post-Incident Report with root cause and defense-in-depth actions after mitigation is complete.”
A coalition of nonprofit organizations sued the top five artificial intelligence companies on Thursday, asking a California court to order a stop to the development of new models that could escape human oversight and cause catastrophe.
The move will most likely keep them stuck in committee until Democrats take control of the lower chamber of Congress in January 2027.
The plaintiffs filed the lawsuit in D.C. federal court on Tuesday, arguing that the structure will permanently overshadow the grave sites of former President John F. Kennedy, former Attorney General Robert F. Kennedy, and other prominent members of the Kennedy family.
Rather, the company’s concern rested primarily in how narrative attacks shaped public perception regarding these races, which, in turn, affect how journalists or pollsters may evaluate voter sentiment.
“Tesla will be winding down production of the Cybertruck at Gigafactory Texas around the middle of next year to make room for increased Cybercab production,” Musk posted on X early Friday morning.
Also by David Zimmermann
Nothing else under this byline is closely related to this article, so these are simply their most recent.
All 161 articles by David Zimmermann →

Topics

ChatGPT LLM Microsoft OpenAI the New York Times

Subjects

OpenAI ORG · 7× Microsoft ORG · 5× the New York Times ORG · 4× ChatGPT ORG · 2× Hecht PERSON · 2× Brent Hecht PERSON · 1× Greg Brockman PERSON · 1× LLM ORG · 1× Nick Turley PERSON · 1× Satya Nadella PERSON · 1×

Narrative

The document lists custom GPTs that can easily extract the user-requested information from paywalled articles, such as “Bypass Paywall” or “Article Reader.” OpenAI does not have a licensing agreement with the New York Times, so the frontier AI lab is accused of severely harming the newspaper’s business model by diverting readers away from its domain.
framing: assertive · carried by 1 article(s) · first seen 2026-09-18
Brent Hecht, Microsoft’s director of applied science, warned about the damaging effects that AI models will have on the internet and the company’s own business.
2026-09-18 · Washington Examiner
OpenAI and Microsoft leaders knew AI may cause ‘doom loop’ hurting entire web · assertive framing

Claims (15 extracted, 3 hedged)

Top executives at OpenAI and Microsoft admitted they knew artificial intelligence poses an “existential threat” to news publications as the New York Times sues the technology companies over copyright infringement. asserted
Times → admit → infringement
Brent Hecht, Microsoft’s director of applied science, warned about the damaging effects that AI models will have on the internet and the company’s own business. asserted
models → warn → internet
“Our AI content strategy has started a ‘doom loop’ that will hurt the performance of our models and the entire web at the same time,” Hecht wrote in an internal document, according to a largely unredacted court filing unsealed Thursday. uncertain
Hecht → start → filing
“It is highly unusual that an end-product threatens the economic foundations of its essential suppliers, but that is the situation we have created for our LLM business with respect to its ‘content supply chain.'” asserted
we → threaten → chain
The New York Times argues chatbots, such as OpenAI’s ChatGPT, should not be able to use its copyrighted content for training because the large language models’ consumption of news articles diverts users from reading their publications. asserted
consumption → argue → publications
The same point was made by Microsoft CEO Satya Nadella, who said chatbots have “substituted” actual journalism by “giving you the information right there on the website on the AI platform versus needing to go to the underlying source.” asserted
chatbots → make → source
Nick Turley, the vice president and head of ChatGPT, similarly said publishers face an “existential threat” from AI chatbots because they “are largely substitutive” and “will get more and more substitutive as they get better.” asserted
they → say → chatbots
Hecht called the rising trend of chatbots accessing news articles for training “an astonishing theft of unprecedented proportions,” adding that perhaps it may be the “largest theft of labor in human history.” uncertain
it → call → history
These past statements from OpenAI and Microsoft leaders will likely bolster the plaintiff’s argument against letting chatbots train on copyrighted works under fair use without permission or payment. asserted
chatbots → bolster → permission
The New York Times has a paywall for its articles, but ChatGPT can evade paying for a subscription. asserted
ChatGPT → have → subscription
OpenAI President Greg Brockman previously bragged about hacking to get behind the news outlet’s paywall in his goal to “scrape” the website, according to the court filing. uncertain
Brockman → brag → filing
The document lists custom GPTs that can easily extract the user-requested information from paywalled articles, such as “Bypass Paywall” or “Article Reader.” OpenAI does not have a licensing agreement with the New York Times, so the frontier AI lab is accused of severely harming the newspaper’s business model by diverting readers away from its domain. asserted
lab → list → domain
However, OpenAI does have licensing deals with other news publishers. asserted
OpenAI → have → publishers
The revelation of AI leaders’ thoughts on the key issues underpinning the legal case comes after the Department of Justice filed a statement of interest supporting the defense made by OpenAI and Microsoft. asserted
Department → underpin → OpenAI
The DOJ argued, in part, that preventing AI models from training on copyrighted material under a “misunderstanding” of fair use “would thwart such creative and scientific progress while hindering American prosperity and economic mobility.” asserted
preventing → argue → prosperity
Give feedback · no sign-in needed

No sign-in needed for now. Your note is sent without a name, so it cannot be answered.

History Support