404 Media maps three routes by which AI floods the web with junk

A weekly podcast from 404 Media places three stories side by side: court records involving OpenAI and Microsoft, AI music inserted onto a real band's Spotify profile and spam sent through the iLands agent platform. Cheap content production connects them, but each case has a different gatekeeper and a different point of failure.

Three stories expose three distinct forms of internet pollution

The first segment points to a 404 Media investigation titled “Doom Loop,” which draws on court records to examine the effects of LLMs on the web and the use of other people's intellectual property. The public podcast description states its conclusion aggressively, while the supporting material sits in a separate article.

The second investigation describes how a reporter managed to place AI generated music on the Spotify page of an existing artist. The third, available in the extended subscriber section, covers iLands, an AI agent platform whose activity 404 Media says is flooding inboxes with spam. The episode is a route map to three separate investigations, not a single proof of a universal theory.

The damage appears in distribution layers that reward volume

A model makes text, music or personalized messages cheaper to produce. Damage requires a distribution system that accepts the material, attaches it to an identity and delivers it to an audience. On Spotify, the key control is the link between a recording and an artist profile. For agent spam, it is sending limits, reputation and operator accountability. For model training, it is data provenance and the legal rules now being contested in court.

That creates an uncomfortable lesson for product teams. Output moderation is insufficient when an agent receives cheap distribution and someone else's identity. Controls must cover the whole chain from generation and account ownership to the platform that amplifies the material.

A headline about killing the internet combines disputes that need separate evidence

The podcast title is intentionally forceful. Intellectual property claims in court, misuse of a Spotify profile and unsolicited email are not interchangeable, however. Each has different victims, rules and standards of evidence.

The public episode page provides a short description and links, rather than the complete case for all three investigations. It is best read as a risk map, with individual claims assessed against their own primary documents.

Platform changes will show up in dispute counts and the cost of abuse

For Spotify, the meaningful signal is whether artist verification improves and disputed attribution is corrected faster. For agent platforms, sending limits, traceable senders and the cost of mass campaigns will matter. For OpenAI and Microsoft, further evidence will arrive through court filings and decisions.

The most useful metric will not be the number of generated songs or messages. It will be how long false material can carry someone else's name and earn money or cause harm before a platform operator intervenes.

Lilith's verdict

A generator can make a song or spam in seconds, but only a platform lends it someone else's name and audience. The internet's waste pipe begins with the operator who fails to notice what is flowing through it.