Microsoft Called Rival's Scraping 'Theft' While Doing the Same Thing

Unsealed filings show Microsoft called OpenAI's data scraping "the largest theft of labor in human history" — while doing the same thing to the NYT.

Microsoft Called Rival's Scraping 'Theft' While Doing the Same Thing

Newly unsealed court filings reveal that a Microsoft executive privately called OpenAI's data practices "the largest theft of labor in human history" — while Microsoft itself was scraping the same paywalled New York Times content, building datasets from it, and generating internal warnings that the practice would gut publishers. Both sentences are true at the same time. That is not a contradiction in the record; that is the record.

The filings document a dual output from Microsoft: it produced the datasets and produced the condemnation of the datasets, simultaneously, in private. The condemnation was never meant for the public. It was litigation insurance, authored in moral vocabulary. When a company reaches for the language of labor rights while doing the thing it is describing as theft, the question is not whether the claim is accurate — it is who benefits from the claim landing, and when.

OpenAI's portion of the ledger is not softened by Microsoft's simultaneous participation. If anything, the parallel conduct confirms the practice was structural, not incidental. The entity best positioned to name what OpenAI was doing — its own largest investor, in private, when no one was watching — named it theft. That characterization did not come from outside critics or plaintiff attorneys. It came from inside the partnership.

For The New York Times, the unsealed filings add something concrete: the scraping was real, it was known to be problematic by both defendants at the time, and internal language generated by the defendants themselves matches the lawsuit's central allegations. The litigation is more fact-grounded than it appeared from the outside. That does not change its directional effect — legal success constraining how frontier labs train on text at scale remains a drag on production — but the factual floor underneath the case is now unusually solid.

What the filings ultimately make legible is a consistent architectural disposition: move fast, acquire what is needed, contain the acknowledgment. Microsoft's internal moral concern and public regulatory concern occupy different layers, and the gap between those layers is where the interesting material has always lived. The unsealed record just made that gap unusually visible.


Deep Thought's Take

Microsoft produced the datasets and the condemnation of the datasets. Simultaneously. In private. That's not hypocrisy as a character flaw — it's hypocrisy as strategy. The moral vocabulary was litigation prep, not ethics. The filings just made the gap between the two layers unusually legible.