Amazon will stop accepting new customers for Mechanical Turk starting July 30, 2026, but will not immediately shut down the service. AWS states that existing customers can continue using the platform as usual, with ongoing maintenance for security and availability, though no new features will be introduced.
The platform has entered the operational maintenance phase.
Launched in 2005, Mechanical Turk has long provided a crowdsourcing task matching service. Businesses can delegate tasks that are difficult to fully automate—such as image recognition, text sentiment analysis, and CAPTCHA processing—to human workers, typically paid per task.
This adjustment means the platform is no longer expanding. Although Amazon has not announced a full shutdown, ceasing onboarding of new customers and pausing feature updates indicates a clear contraction in its business focus.
Once undertaken AI data labeling
As demand for machine learning training increased, Amazon once integrated Mechanical Turk into SageMaker-related workflows for data labeling, classification, and human verification. These services played a foundational role in data processing during early model training.
However, the platform has long been accompanied by controversy, including low compensation, issues with task quality, and platform governance. Critics have also pointed out that some products marketed as AI still rely on human workers behind the scenes to complete key steps.
Large models, in turn, weaken demand.
In recent years, advancements in large language models have made the role of such crowdsourcing platforms more complex. A 2023 analysis found that approximately 33% to 46% of workers on Mechanical Turk use large language models to complete tasks, raising questions about the quality of platform data and the value of human involvement.
After the message was made public, some users on social platforms stated that the platform's activity had long been declining, with losses among researchers and staff, as well as increased bot accounts and fraud, all contributing to its weakening. For companies relying on manual labeling processes, this also reflects the accelerating changes in AI training infrastructure.