Automation

Amazon shutters Mechanical Turk after 21 years

The crowdsourced data labeling platform that Jeff Bezos dubbed 'artificial artificial intelligence' will close September 30, 2026.

Omega Editorial· August 26, 2026· 3 min read

Amazon will shut down Mechanical Turk, its crowdsourced work platform, on September 30, 2026, ending a 21-year run that helped define the gig economy and early data labeling industry.

The company announced the closure Tuesday in a notice posted to the MTurk website, stating it had assessed the program and decided to discontinue the service. Amazon did not provide specific reasons for the shutdown.

The original 'artificial artificial intelligence'

Launched in 2005, Mechanical Turk connected workers with simple digital tasks—dubbed "Human Intelligence Tasks"—that computers struggled to complete but humans could easily handle. These included labeling data, transcribing audio and video, and completing surveys, typically paying a few cents per task.

Jeff Bezos called the service "artificial artificial intelligence," a reference to its approach of using human cognition to solve problems that automated systems couldn't tackle. The name itself came from an 18th-century chess-playing automaton that appeared mechanical but was secretly operated by a human chess master hidden inside.

Amazon initially developed the platform to help label vast amounts of data across its e-commerce operations, according to The New York Times. At its peak, MTurk served more than 500,000 workers, known as "turkers."

Decline amid AI advances and new competition

The platform had been "in decline" in recent years, according to Krista Pawloski, a data worker and organizer with Turkopticon, an employee advocacy group. As AI models advanced rapidly, they began handling tasks that once required human judgment. Meanwhile, specialized data labeling startups including Scale AI, Mercor, and Prolific emerged to serve the booming demand for training data.

Amazon attempted to position MTurk as a data annotation resource for its SageMaker machine learning service, promoting the workforce's 24/7 availability. But the company appeared to reduce investment in improving the platform even as competitors gained ground, Pawloski said, prompting many workers to migrate to other services.

Quality concerns also surfaced. A 2023 study by Swiss academic researchers found that up to 46% of Mechanical Turk workers were using AI models to complete their assigned tasks.

Impact on workers and businesses

Amazon stopped accepting new MTurk customers last month, a move workers interpreted as foreshadowing the closure. Some insurance and travel companies still rely on the platform for data work and are now searching for alternatives, Pawloski said.

For workers, the shutdown eliminates a source of flexible, remote income. Pawloski began using MTurk in 2008 during maternity leave and transitioned to full-time turking in 2012 after losing her job, allowing her to care for her son with special needs while earning income. She noted that some workers still depend on the platform as their primary source of employment.

Why it matters

Mechanical Turk's closure marks the end of a pioneering experiment in distributed human computation that helped lay groundwork for today's data labeling industry. Its demise reflects how rapidly AI capabilities have evolved—tasks that required human intelligence two decades ago are now increasingly automated. The shutdown also highlights consolidation in the data annotation market, where specialized platforms have displaced Amazon's generalist approach. For the thousands of workers who relied on MTurk's flexible, remote opportunities, the closure underscores the precarity inherent in gig-based data work.

The details were first reported by CNBC.

#amazon#mechanical turk#data labeling#gig economy#crowdsourcing#ai training data

This is an original analysis by the Omega editorial team. Source reporting: AI Watch.

Want systems like this working for your business?

Book a Call

More in Automation

Automation· 3 min read

Why Expense Claims Are Fintech's Hardest Automation Problem

The technical challenge isn't reading receipts—it's encoding the messy eligibility rules that determine whether a claim gets paid.

Via Automation Watch · Aug 26, 2026
Automation· 4 min read

Two Seattle AI Startups Take Different Paths to Weed Control

TerraClear maps individual weeds for precision spraying while Carbon Robotics deploys laser-equipped robots to eliminate them without chemicals.

Via AI Watch · Aug 26, 2026
Automation· 3 min read

Tesla Drivers Filmed Sleeping Behind the Wheel in 43 Videos

A growing number of Tesla owners are documented napping while their vehicles operate in autonomous driving mode, raising safety concerns.

Via The Verge · Aug 25, 2026