OpenAI's copyright fears were named, HN specifically

Newly unsealed court filings in the Authors Guild v. OpenAI case quote an OpenAI researcher expressing worry that using data from a 'sketchy Russian website' would look bad if it 'showed up on HN.' The thread on this is now, ironically, on HN. Commenters pushed back on the framing, noting the site in question was Bibliotik, a library, not some criminal operation, but the more damaging detail is that internal concern existed and training continued anyway.

The pattern here: OpenAI's internal communications are now becoming evidence in litigation, and the reputational calculus the company made (weigh the PR risk, proceed) is being litigated in public. This is not just a legal story. It's a trust story about how the foundational training data decisions were made.

Counterpoint in the thread: some commenters argued the Authors Guild is itself a lobbying organization selectively presenting evidence. That may be true, but it doesn't change what the researcher said.