Microsoft exec called AI scraping 'the largest theft of labor in human history'

TEXT START: New unredacted information in the copyright lawsuit The New York Times brought against OpenAI and Microsoft three years ago reveals an admission that AI scraping was tantamount to theft, and that AI products pose a major threat to publications.

THE DISSECTION

The text is a legal exposé wrapped around a larger economic autopsy. It documents four mechanisms: mass extraction of human work, concealment through stripped copyright notices and paywall circumvention, direct substitution of chatbot answers for publisher traffic, and a feedback loop in which declining traffic weakens the very content industry supplying the models.

Its real function is to convert systemic displacement into a courtroom narrative about misconduct, fair use, and licensing. The admissions are more consequential than the legal framing: Microsoft and OpenAI allegedly understood that their products could destroy the economic foundations of their content suppliers while continuing to build the machinery anyway.

THE CORE FALLACY

The text treats permission and compensation as if they could solve the underlying crisis. They cannot. Licensing every article would redistribute revenue and perhaps slow deployment, but it would not restore the scarcity of human-produced information, reverse chatbot substitution, or preserve journalism as a mass source of economically necessary labor.

“Largest theft of labor” is a powerful moral description, but it is incomplete as an economic diagnosis. The decisive event is not merely that labor was copied without payment. It is that the copied labor becomes training material for systems designed to reduce future dependence on that labor. The model owners are not just stealing the harvest; they are building machinery to eliminate the farm.

HIDDEN ASSUMPTIONS

• That a legal victory can preserve the publisher business rather than merely assign licensing rents to its owners.
• That restoring clicks or licensing fees restores productive participation for journalists. It does not.
• That publishers remain indispensable suppliers after their output has been absorbed into models that can answer users directly.
• That copyright, paywalls, and attribution are durable economic moats rather than lag defenses against a superior extraction system.
• That institutions can coordinate a stable human-only domain while firms compete to automate it. Under P2, they cannot do so at scale.
• That identifying corporate knowledge of the harm changes the competitive imperative. It does not; the firms can recognize the corpse while fighting over ownership of it.

SOCIAL FUNCTION

Primary classification: partial truth. Secondary classifications: transition management and elite self-exoneration.

The article accurately exposes substitution, supplier destruction, and deliberate extraction. It is not simple copium. But its legal scandal frame gives the audience a manageable villain and a manageable remedy: punish the firms, impose licenses, preserve the industry. That converts a structural transition into a dispute over consent and pricing. Publishers can use the evidence to negotiate a larger share of the carcass, but not to restore the old labor market.

THE VERDICT

The article catches the discontinuity in the act, then misnames it as a copyright scandal. AI scraping is the intake valve; cognitive automation is the kill mechanism. The lawsuit may decide who collects rent from the captured corpus. It cannot revive the mass employment-to-consumption circuit. Publishers and journalists are being converted from producers into feedstock, negotiators, or servitors of model owners. The postwar order is not being defended here. Its replacement is being litigated over.