Previously redacted LLM “doom loop” material surfaces in unsealed filing

Internal material quoted in an unsealed copyright filing

404 Media reports that an unsealed New York Times filing seeking summary judgment cites documents and testimony from Microsoft and OpenAI personnel that had previously been sealed or redacted. According to the report, the filing includes a Microsoft document describing an AI-content “doom loop,” in which AI products draw on creator content while reducing traffic to the sources that supply it. The report also attributes to the filing allegations about LLM training material, a claimed more-than-90% fall in Bing clicks to news sites, and an alleged OpenAI workaround for the New York Times paywall.

The cited Microsoft and OpenAI material had reportedly been sealed or redacted from public view in the New York Times copyright case.

A filing seeking summary judgment was unsealed with previously redacted material that the report characterizes as internal acknowledgments of risks to publishers and the web from LLM products.

The reported disclosure changes what is publicly visible in a major copyright dispute: arguments over LLM training and effects on publishers are presented as including internal company material, rather than only outside criticism. The characterization and underlying evidence remain unverified here.