Microsoft director: AI scraping 'the largest theft of labor in human history'

TEXT START: The NYT argues that OpenAI and Microsoft infringed upon its copyright over thousands of news articles.

The Dissection

This is not fundamentally a copyright story. It is an exposure of AI’s supply-chain logic: models ingest publishers’ labor, generate substitutes for publisher output, and reduce the traffic that finances the original labor. The quoted Microsoft and OpenAI communications show that senior personnel understood both the extraction and the substitution risk.

The article then compresses a systemic economic rupture into a courtroom dispute over fair use, paywalls, licensing, and summary judgment. That framing is legally relevant but strategically narrow. The evidence comes through the NYT’s legal brief, while some documents remain sealed or redacted; it demonstrates alleged internal awareness, not final liability.

The Core Fallacy

The article’s central error, under Discontinuity Thesis mechanics, is treating proof of unauthorized use or a licensing settlement as a potential restoration of the old publisher economy.

The decisive event is not merely that labor was “stolen.” It is that the labor is being made less necessary as an ongoing human service. Licensing can determine who collects rent on the input. It cannot necessarily restore readership, referral traffic, pricing power, or employment once users accept AI-generated substitutes. The article identifies the wound, then mistakes legal compensation for structural recovery.

Hidden Assumptions

• Courts can enforce durable licensing boundaries against rapidly scaling AI systems.
• Compensation for scraped material will be large and stable enough to sustain current newsroom economics.
• Links, citations, or attribution will preserve user demand, despite the quoted admission that users may not click.
• Publishers will remain indispensable after their content is converted into model capability.
• Legal recognition of market harm will translate into bargaining power rather than a temporary transfer payment.
• Retraining models or licensing paywalled material would preserve the publisher’s role rather than merely formalize its conversion into an upstream data supplier.

Social Function

Primary classification: partial truth and transition management.

The article reveals a real contradiction: AI firms depend on content suppliers while building products that can eliminate those suppliers’ distribution and revenue. But by centering the dispute on alleged scraping and fair use, it encourages the audience to believe that proper licensing would solve the problem. That is the anesthesia. The deeper issue—AI’s substitution of economically necessary human cognitive work—remains largely outside the legal frame.

It also functions as prestige signaling and limited elite self-exoneration: executives appear candid about the danger, but acknowledging the machine’s effects is not the same as relinquishing the machine’s advantages.

The Verdict

This is a confession from inside the transition. Microsoft and OpenAI appear to understand that their systems consume publishers’ labor and then attack the economic channel that made that labor viable.

The lawsuit may produce damages, licenses, or a temporary moat. It cannot, by itself, restore the mass employment-to-consumption circuit or make human publishers economically indispensable. Under the Discontinuity Thesis, this is carcass management: publishers are fighting over compensation for becoming upstream data suppliers while AI absorbs the downstream service.

The article does not prove the entire P1–P3 framework. It does provide a direct miniature of it: extraction, substitution, supplier collapse, and legal lag. The law may slow the burial. It does not resurrect the body.