Opinion

Microsoft Exec Privately Called AI Scraping 'the Largest Theft of Labor in Human History'

Newly unsealed court filings show Microsoft internally acknowledged that its training data practices wouldn't hold up to scrutiny — a sharp contrast to its public 'Responsible AI' messaging.

Font size

The incident: a court filing cracks things open

TechCrunch reported on September 17 [1] that a batch of previously redacted court documents has been unsealed in ongoing AI copyright litigation. The filings contain internal communications in which a Microsoft executive referred to AI scraping as "the largest theft of labor in human history."

The story hit 230 points and 154 comments on Hacker News — solid engagement for a piece of legal news.

My first reaction after reading it: the weight of that statement isn't in the word "theft" — it's in "labor."

Why this lands so hard

Microsoft has spent the last two years leaning hard on its Responsible AI brand. Satya Nadella has spoken publicly about Microsoft's AI principles, and Azure's compliance docs repeatedly emphasize that training data must be clean and IP-compliant.

Then an internal exec casually calls it the "largest labor theft."

This isn't a PR slip. It's a cognitive dissonance. When a company tells the public "our training data is responsible" while internally framing the same activity as "theft," it means the decision-makers already know the score: the practice can't withstand real scrutiny, but the business has to do it anyway.

The awkward part: this admission didn't come from outside critics. It came from their own people.

"Theft of labor" — a deliberate word choice

Look closely at the phrasing: "theft of labor," not "theft of content" and not "copyright infringement."

  • Copyright infringement — a technical legal question. You can argue "fair use."
  • Theft of content — vague, with a moral tint.
  • Theft of labor — points straight at people, at the time, skill, and judgment creators put in.

The third framing reframes AI training data from "legal gray zone" to ethical problem: you didn't take data; you took someone's work. That's a qualitative judgment you can't wave off with legal hand-waving about "transformative use."

That's why some HN commenters said that if a jury sees this, Microsoft is basically done. It's not hyperbole. In a legal context, "admitting theft" and "admitting possible infringement" are evidence of completely different magnitudes.

The other big players are in the same boat

Microsoft isn't the only one being sued.

  • OpenAI: facing a lawsuit from The New York Times over unauthorized use of news content to train GPT.
  • Google: targeted by multiple class actions; the source of Gemini's training data is equally opaque.
  • Meta: forced to use euphemisms like "the Books3 dataset" in internal documents after training LLaMA on LibGen (a shadow library).

But in all those cases, the defendants' public line is some version of our training data is legitimate. What's different here: Microsoft wasn't outed by external accusations. It was caught by its own internal documents.

The distance between a court filing and a press release is the real gap between an AI company's actual ethical position and its public one.

What this means

Put this alongside the litigation from the past 18 months:

  1. Late 2023: OpenAI and The New York Times negotiations collapse; lawsuit filed.
  2. 2024: A wave of creator class actions enter substantive proceedings.
  3. 2025–2026: Settlement amounts start surfacing — many undisclosed.
  4. September 2026: Microsoft internal filings unsealed; the word "theft" officially enters the court record.

The trend: litigation is moving from the "allegation" phase into the "discovery" phase. Discovery means both sides hand over internal documents, and what's in those documents is almost always more honest than the press releases.

Editor's take

This isn't just another copyright dispute. What's really being exposed is a less comfortable fact: AI company executives know the legality of their training data doesn't hold up, but commercial pressure forced them to do it first and ask questions later.

A few things worth watching:

  • Whether other plaintiffs cite these filings. Once evidence goes public, it stops being about just one case.
  • Whether Microsoft tries to isolate the statement — pin it on one exec rather than company policy. But "the largest theft of labor in human history" doesn't sound like something a junior employee mutters off the cuff; it reads like a leadership-level judgment.
  • Whether the EU AI Act and various US state AI laws accelerate. The law usually lags technology, but exposed filings force legislators to take a position.

A contrarian read: in the short term, this won't move Microsoft's stock much — the market has already priced in ethical controversy from AI companies. In the medium term, though, it will shift how enterprise customers feel about "using Microsoft AI services." B2B buyers don't care whether AI is a genius; they care whether vendor legal risk is contagious. The moment a major customer refuses to renew an Azure contract because "your vendor used data we can't touch," the entire B2B AI market reshuffles.

What I don't know yet: who the exec actually is, or the context of the quote (email? meeting notes? Slack?). If it came out of a formal decision-making meeting, the damage goes up another notch.


References

[1] TechCrunch, "Microsoft exec called AI scraping 'the largest theft of labor in human history,' new unredacted filings reveal", 2026/09/17. https://techcrunch.com/2026/09/17/microsoft-exec-called-ai-scraping-the-largest-theft-of-labor-in-human-history-new-unredacted-filings-reveal/

Source: https://techcrunch.com/2026/09/17/microsoft-exec-called-ai-scraping-the-largest-theft-of-labor-in-human-history-new-unredacted-filings-reveal/