Amazon has posted a closure notice on the official Mechanical Turk site confirming that its crowdsourcing marketplace will permanently close on September 30, 2026. The decision ends the service that matched businesses and developers with a distributed human workforce for small digital tasks, which Amazon termed Artificial Artificial Intelligence.
“We regularly evaluate our programs, tools, and services and make adjustments based on those assessments,” the notice reads. “Following an assessment, we’ve made the decision to close Amazon Mechanical Turk, effective September 30, 2026.”
Shutdown Mechanics for Workers and Requesters
Amazon’s closure FAQ lays out the wind-down for both sides of the marketplace. Workers are directed to verify payment preferences inside the MTurk Worker Portal, including a payment method, payment frequency, and transfer schedule, and payments after closure will be distributed based on each worker’s existing schedule. All completed and approved Human Intelligence Tasks, the platform’s unit of work known as HITs, will be paid out during the worker’s normal scheduled disbursement, and MTurk will still distribute tax-related documentation through its established processes and deadlines.
HIT submission closes on September 30, 2026, and all remaining unsubmitted HITs will automatically expire. Requesters then have 30 days under the platform’s standard approval policy to approve or reject completed work, until October 30, 2026; if no action is taken, the HITs will be auto approved. Requesters can award bonuses to workers up until that same October 30, 2026 date.
Requesters are instructed to verify their payment information within the MTurk Requester Portal so refunds can be processed in a timely manner. Pre-paid Requester balances will be refunded in full within 30 days, and customers will receive a final bill from MTurk in their upcoming AWS billing cycles. Transaction history will remain available until January 28, 2027.
Reach Into SageMaker Ground Truth and Augmented AI
The closure extends beyond the standalone marketplace. According to the FAQ, it also applies to SageMaker Ground Truth and Amazon Augmented AI, where the Amazon Mechanical Turk Worker type will no longer be available when creating labeling jobs or human review workflows as of September 30, 2026. Teams that built machine learning data pipelines on Amazon’s own crowd workforce will need a different worker source for those tools after that date.
The Marketplace Amazon Built
The service takes its name from an 18th-century chess machine. Amazon’s FAQ recounts that in 1769, Hungarian nobleman Wolfgang von Kempelen built a mechanical chess-playing automaton that defeated nearly every opponent it faced, touring Europe and confounding challengers including Benjamin Franklin and Napoleon Bonaparte. Kempelen opened the cabinet’s doors to reveal gears, cogs, and springs and convinced audiences he had built a machine that made decisions using artificial intelligence; the secret was a chess master concealed inside.
Amazon applied the same idea to software. The company’s FAQ states that humans still significantly outperform the most powerful computers at completing such simple tasks as identifying objects in photographs, something children can do before they learn to speak. MTurk let Requesters integrate what Amazon called Artificial Artificial Intelligence directly into their applications by making requests of humans, using the web interface or web services API to submit tasks, approve completed work, and incorporate the answers. When using the API, the transaction looked like any remote procedure call; in reality, a network of human Workers fueled the system by completing tasks and receiving rewards.
Amazon came up with the idea to help solve specific internal data processing problems that required human judgment and intelligence, then decided to expose it as an external web service so Requesters could build applications with human intelligence as a core component. The marketplace pitched a global, on-demand, 24×7 workforce for use cases including machine learning development, business process outsourcing, data validation, survey participation, and content moderation. Common examples included moderation of web and social media content, categorization of products or images, and collection of data from websites.
For machine learning workflows, MTurk positioned itself as a way to collect and annotate the massive amounts of data required to train models, and as a human-in-the-loop option where human feedback helps validate and retrain a model. One cited example was drawing bounding boxes to build high-quality datasets for computer vision models, work described as too ambiguous for a purely mechanical solution and too vast for even a large team of human experts.
Pricing and Platform Model
Under the platform’s published pricing, the price a Requester pays for a HIT has two components: the reward paid to Workers, which the Requester sets, plus a fee paid to MTurk based on that reward. The standard fee is 20% of the reward and bonus amount, with an additional 20% fee on the reward for HITs carrying 10 or more assignments and a minimum fee of $0.01 per assignment or bonus payment. Using the Masters Qualification, awarded to workers who demonstrated excellence across a wide range of HITs under ongoing statistical monitoring, adds a fee of 5% of the reward, while Premium Qualifications carry an additional per-assignment fee starting at $0.05.
Requesters outside the United States were accepted from a list of countries across the Americas, Africa, Asia and the Pacific, and Europe. Workers could complete tasks on their own schedule with a computing device connected to the Internet and were required to be at least 18 years old, collecting earnings through scheduled transfers once a Requester approved their work.
Final Timeline
The September 30, 2026 closure date governs the end of HIT submissions and the removal of the Mechanical Turk Worker type from SageMaker Ground Truth labeling jobs and Amazon Augmented AI human review workflows. Bonuses and HIT approvals run through October 30, 2026, with unactioned submissions auto approved after the 30-day window. Requester refunds are due within 30 days, and transaction history stays accessible through January 28, 2027.
Credit: Source link
