Amazon just killed Mechanical Turk. The platform that trained AI just got replaced by AI.

Amazon is shutting down Mechanical Turk on September 30.

21 years. 500,000 workers. The platform Jeff Bezos called "artificial artificial intelligence" — humans doing tasks too hard for computers — is dead.

Because the models it trained can now do the work themselves.

The irony is structural, not poetic. A 2023 EPFL study found 33-46% of MTurk workers were already using ChatGPT to complete their tasks. Humans were being paid to relay machine guesses back to the machine.

Now Scale AI, Mercor, and Prolific have absorbed the market. Amazon shut off new customer signups on July 30. The notice to existing customers was four sentences flat.

Here's what your CISO and CHRO need to hear simultaneously.

MTurk wasn't just a gig platform. It was the annotation layer underneath half the AI models your enterprise deployed. Every training dataset, every fine-tuned classifier, every evaluation benchmark — a chunk of that human signal came from workers earning 20 cents per task.

Those workers are now unemployed. And the annotation vendors replacing them cost 3-5x more per labeled unit because they run AI-assisted pipelines with human oversight, not raw human labor.

Your data labeling budget just went up. Your vendor dependency just got deeper. And the workforce that powered the AI revolution just got left behind by it.

Audit your annotation supply chain today. If you're still routing through legacy crowdsourcing platforms, you're already behind. The replacement vendors are Scale AI, Mercor, and Prolific — and they're not cheap.

SOURCE: https://www.cnbc.com/2026/08/25/amazon-service-that-jeff-bezos-called-artificial-ai-is-shutting-down.html
VERIFIED: CNBC (primary), AI Weekly, GuruFocus, SDxCentral
SIGNAL: The platform that trained the models that replaced it. This is the clearest case study of AI cannibalizing its own labor infrastructure — and your annotation budget is about to feel it.