Morning Edition · №
Technology · AI & Labor SEATTLE

Amazon Will Close Mechanical Turk, the Marketplace That Trained the AI Era

AWS says it will shut down the 21-year-old crowdwork platform on September 30, alongside two related data-labeling tools, as AI systems and specialized startups take over the work it once managed.

Amazon Will Close Mechanical Turk, the Marketplace That Trained the AI Era
An Amazon Web Services data center. Mechanical Turk was an AWS-operated marketplace rather than a physical facility, but the service falls under the same cloud division. — Photograph: Amazon Web Services, official press kit
SHARE X f in ⧉

Amazon Web Services will close Mechanical Turk, the online marketplace where businesses have paid people small sums to complete short digital tasks since 2005, effective September 30. The company confirmed the closure this week, telling customers in a statement that "following an assessment, we've made the decision to close AWS Mechanical Turk, effective September 30, 2026." SageMaker Ground Truth and Amazon Augmented AI, two related services built for routing data-labeling work to human reviewers, are being retired the same day.

Mechanical Turk — nicknamed "MTurk" by its users and once described by Jeff Bezos as "artificial artificial intelligence" because it let companies request human judgment through an API — began as a general-purpose task marketplace, handling everything from CAPTCHA-solving to survey responses. It was repositioned in 2018 around a narrower purpose: supplying the labeled training data that early machine-learning and computer-vision systems needed to improve. Amazon had already stopped accepting new customers on the platform on July 30, a step widely read at the time as a prelude to a full wind-down.

The workforce AI training no longer needs

The platform's retirement closes a chapter research had already begun to complicate. A widely cited 2023 study found that as many as 46% of Mechanical Turk workers appeared to be using large language models themselves to complete the very tasks meant to train those models, undermining the reliability of the data Amazon's customers were paying for. At the same time, a newer generation of data-labeling companies — Scale AI, Mercor and Prolific among them — has pulled workers toward higher-value assignments, recruiting programmers, scientists and domain experts for the kind of reinforcement-learning and evaluation work that today's frontier AI labs now depend on far more than simple microtasks.

Amazon's own account, reported by Tech Startups, points to a broader reallocation of resources rather than a single trigger: the company invested less in Mechanical Turk as competing platforms gained ground, and modern AI systems now handle image classification, transcription and basic data processing tasks that once required a human worker on the other end of a HIT — Amazon's term for a Human Intelligence Task.

The closure leaves existing customers with roughly five weeks to migrate workflows elsewhere, and ends, at least at Amazon, the era in which a company built specifically to supply human labor marketed itself as an AI service. That framing was novel in 2005; two decades later, with AI systems doing much of the work Mechanical Turk workers used to do, and the workers who remain increasingly leaning on AI to do it, the platform's original premise had largely inverted.

SHARE THIS ARTICLE X Facebook LinkedIn Copy link
Claire Fontaine · Technology & Regulation Correspondent

Reports on technology and its regulation for UBStandard, with a focus on Brussels, AI policy and Europe's digital economy.

[email protected]
Related coverage Front page →