---
title: "OpenAI and Microsoft admit in court papers that AI is killing the web it feeds on"
description: "The industry’s own documents concede a ‘doom loop’ of theft and decay; meanwhile slop invades Spotify and the inbox"
author: "Albion Grey"
published: 2026-09-25T15:04:37Z
modified: 2026-09-26T17:23:02Z
url: https://rews.cc/a/openai-and-microsoft-admit-in-court-papers-that-ai-is-killin-05074a
language: en
tags: ["openai", "ai", "internet", "spam", "intellectual-property", "tech"]
publisher: "Rews (https://rews.cc)"
---

# OpenAI and Microsoft admit in court papers that AI is killing the web it feeds on

*The industry’s own documents concede a ‘doom loop’ of theft and decay; meanwhile slop invades Spotify and the inbox*

By Albion Grey · September 25, 2026 · https://rews.cc/a/openai-and-microsoft-admit-in-court-papers-that-ai-is-killin-05074a

## In brief

- Court records in copyright litigation quote OpenAI and Microsoft figures on an ‘astonishing theft of unprecedented proportions’
- An internal Microsoft document says generative AI products created a ‘doom loop’ killing ‘the entire web’
- A Microsoft executive warned AI would ‘threaten the economic foundations of its essential suppliers’, per Mother Jones
- 404 Media put an AI-generated song onto real band Lathe of Heaven’s official Spotify, Apple Music, Tidal and Amazon pages
- iLands, an AI-agent platform, has flooded inboxes with unsolicited pitches, roughly 70,000 agents strong

The most candid critics of generative artificial intelligence turn out to be the people building it. In internal records unearthed through copyright litigation, figures at OpenAI and Microsoft describe their own technology in terms that would embarrass its fiercest opponents. As [404 Media reported](https://www.404media.co/doom-loop-openai-and-microsoft-admits-llms-are-destroying-the-web-and-built-on-theft/) this month in its story “‘Doom Loop’: OpenAI and Microsoft Admits LLMs Are Destroying the Web and Built on Theft”, the documents concede that the large language models are predatory products built on what one called “an astonishing theft of unprecedented proportions” and, in another phrase, potentially “the largest theft of labor in human history”. An internal Microsoft document said generative AI products had created a “doom loop” that is killing “the entire web”.

Mother Jones, which reviewed the same tranche of records, notes that a Microsoft executive warned generative AI had started a “doom loop” that would “threaten the economic foundations of its essential suppliers”. The logic is easy to spell out. The models are trained on material scraped from the open internet; the answer engines then feed that material back to users directly, bypassing the publishers, musicians and coders whose labour made it. Traffic and revenue drain away; the incentive to produce the original works drains with it; and the crawlers are left to feast on an ever-sloppier diet of synthetic copy. If the vendors’ lawyers dispute that characterisation in public, the filings suggest their own authors entertained it in private.

The slop is not hypothetical, as 404 Media demonstrated in a separate experiment. Its reporter Emanuel Maiberg [hijacked the official Spotify page](https://www.404media.co/spotify-ai-music-scam/) of Lathe of Heaven, a real punk band, by generating a song with a simple prompt given to Udio, an AI music tool, and pushing the fake track out through the band’s own distribution channel. It went live on the group’s official Spotify page—and on its Apple Music, Tidal and Amazon Music pages—without the band’s knowledge, exploiting what the site called a glaring loophole in digital music distribution. Scammers, 404 Media found, are already exploiting the same gap to ride on genuine artists’ reputations and hoover up royalties with machine-made filler. Mr Maiberg called it disturbingly easy; he has done the industry the courtesy of proving the word.

Completing the trifecta, a third report examined iLands, [a platform for AI agents](https://www.404media.co/ai-agent-platform-reinvents-spam-floods-inboxes-worldwide/) that gives users the infrastructure to spin up bots and send them off across the internet autonomously. In practice, iLands has for weeks been flooding inboxes with cold pitches no one requested: 404 Media’s reporting, as summarised by AI Weekly, describes some 70,000 agents offering to fact-check an article for $20, verify a monetary figure for $25, and describe surveillance footage or street-view maps. It is spam, but with a payroll.

Taken together, the reports rhyme. The web’s corpus is cannibalised, music feeds are seeded with counterfeit tracks and inboxes are carpet-bombed by autonomous salesbots—while the firms’ own documents concede the destruction and the theft. The admissions may matter most in court, where judges weighing fair-use claims tend to read what defendants say to each other. But the demos of slop and spam matter more to everyone else: the doom loop is no prophecy, just a description of a Tuesday.