Amazon is shutting down the human workforce that spent twenty-one years secretly training AI, because AI finally made their jobs obsolete.
Here's the full irony, via TheNextWeb's reporting.
What It Was
Mechanical Turk launched in 2005, paying people a few cents at a time for "Human Intelligence Tasks" -- small jobs like labeling data or transcribing audio that computers couldn't do. Jeff Bezos famously called it "artificial artificial intelligence." Amazon announced the closure on August 26, 2026; it stopped accepting new customers on July 30 and closes for good on September 30.
The Twist
"A 2023 study by academic researchers in Switzerland found that up to 46% of Mechanical Turk workers were using AI models to complete their tasks." The very platform proving humans were needed was quietly being outsourced to AI by its own workers.
Why That Matters
"The service existed because some jobs were easy for people and hard for machines. That gap is what the study was measuring, and what has closed."
The Human Cost
Organizer Krista Pawloski has been on the platform since 2008 -- she started turking during maternity leave, then went full-time in 2012 after losing another job, drawn to work she could do remotely while caring for her son, who has special needs. She's organized with the worker advocacy group Turkopticon since 2019, and describes MTurk work at its best as offering "meaningful income with flexible hours," comparable to ride-hailing but doable from home. Her read on why this happened isn't purely "AI took the job" -- it's also neglect: "MTurk had been in decline for years. Amazon put fewer resources into improving it while rivals arrived, and many turkers moved to competing platforms." Still, for the workers who stayed: "There's some people that still pretty much still do it full time. They're concerned now." Insurance and travel companies still relying on MTurk for data work have just 35 days from the announcement to find alternatives.
Where the Work Actually Went
The work itself hasn't disappeared -- newer platforms like Scale AI, Mercor, and Prolific have absorbed the demand for training data. What's changed is who runs that market, and what they pay for it: these platforms pitch themselves on "curated workforces and expert annotation" at meaningfully higher rates than MTurk's cents-per-task model, a shift from anonymous piecework toward smaller pools of more specialized, better-paid labor.