Amazon is pulling the plug on Mechanical Turk, the 21-year-old crowdsourced work platform that founder Jeff Bezos once famously described as "artificial artificial intelligence." The service will close for good on September 30, 2026, according to a notice posted Tuesday on the Mechanical Turk website.

The shutdown closes a remarkable chapter in the history of computing — a service built to fake machine intelligence with human labor that ended up supplying the very data used to train the genuine article. For more context on this story, see our ongoing breaking AI news.

"We regularly evaluate our programs, tools and services and make adjustments based on those assessments," the company's notice read. "Following an assessment, we've made the decision to close AWS Mechanical Turk, effective September 30, 2026."

How Mechanical Turk Worked

Launched in 2005, Mechanical Turk matched legions of online workers with simple digital jobs known as "Human Intelligence Tasks," or HITs, that typically paid out a few cents each. The tasks included labeling data, transcribing audio or video, and answering surveys — work that was trivially easy for humans but remained stubbornly difficult for the computers of the era.

The platform's name was a deliberate historical wink. It referenced the original Mechanical Turk, an 18th-century chess-playing automaton that awed European audiences before it was revealed to be a hoax, secretly controlled by a human chess master hidden inside the cabinet. Bezos embraced the parallel, calling the service "artificial artificial intelligence" — real people doing the work that machines could not.

Amazon originally conceived the platform to help label reams of product data on its sprawling webstore, The New York Times reported. At its peak, the service served more than 500,000 workers, known colloquially as "turkers," according to figures cited by CNBC.

Eclipsed by the Technology It Mimicked

The irony of Mechanical Turk's demise is hard to miss. The platform spent two decades renting out human cognition by the task — and the datasets those humans produced helped train the large language models that ultimately made much of that piecework obsolete.

A 2023 study from academic researchers in Switzerland found that up to 46% of Mechanical Turk workers were already using AI models to complete their tasks, blurring the line between the human intelligence the platform was designed to sell and the machine intelligence it helped create.

Meanwhile, a new generation of data startups — including Scale AI, Mercor and Prolific — moved aggressively into the market, recruiting workers specifically to train and evaluate AI models, often at better rates and with more modern tooling. Amazon itself tried to reposition Mechanical Turk as a data annotation source for its SageMaker machine learning service, with an AWS webpage promising that "workers are available 24 hours a day, 7 days a week."

Warning Signs and a Slow Decline

The final closure did not come out of nowhere. Last month, Amazon said it would stop accepting new Mechanical Turk customers, a move many workers interpreted at the time as a sign that the service was living on borrowed time.

Krista Pawloski, a data worker and organizer with Turkopticon, an advocacy group for data workers, told CNBC the platform had been "in decline" for years. She said Amazon appeared to invest fewer resources in improving Mechanical Turk just as competing data labeling services arrived, prompting many turkers to migrate to rival platforms.

The closure leaves real dependencies exposed. Pawloski said some insurance and travel companies still rely on Mechanical Turk for data work and are now scrambling to find alternative platforms.

A Lifeline for Some, Now Gone

For many workers, Mechanical Turk was more than a curiosity — it was income. Pawloski started turking in 2008 while on maternity leave, then shifted to the work full time after losing her job in 2012, using the flexible, remote income to help care for her son with special needs.

"There's some people that still pretty much still do it full time," she said. "They're concerned now."

Amazon did not immediately respond to a request for comment on the closure, CNBC reported.

The End of an Era in Human-Powered Computing

Mechanical Turk occupies a unique place in the story of artificial intelligence. It arrived in 2005 — years before deep learning reignited the AI field — as a pragmatic admission that some problems still needed a human in the loop. It pioneered the crowdsourced labor model that would later power everything from content moderation pipelines to the human feedback systems used to align modern chatbots.

That the platform is closing in the same decade its output helped train is a neat, if bittersweet, bookend. The fake chess robot of the 1770s was retired once the trick was exposed. Amazon's version survives a little over two decades — until the machines, fed by millions of tiny human tasks, finally caught up.

For workers, the question now is where the remaining crowdwork demand goes next — and whether the platforms that absorbed it, built explicitly for the AI era, will treat the humans in the loop any better than the one that started it all.

---

Stay Ahead of AI

Get the latest AI news, analysis, and breakthroughs — all in one place.

Read more AI news →