The Amazon will stop receiving new clients for the Mechanical Turk from July 30, 2026 but will not close the service immediately. AWS indicates that existing clients are still available and that the platform will continue to maintain security and availability but will not introduce new functionality.
The platform enters the maintenance operational phase
Machanical Turk came online in 2005 and provided a long-term crowd-sourcing package. Enterprises can hand over tasks that are difficult to fully automate to manual processing, such as photo identification, text emotional judgement and authentication code processing, and tasks are usually billed on a case-by-case basis.
This realignment means that the platform will not continue to expand. Although the Amazon did not declare a full cut-off, the cessation of the absorption of new clients and the simultaneous suspension of the upgrading of functions showed a marked contraction in the focus of its operations.
Once assumed AI data label
As the need for machine-learning training has risen, Amazon has integrated Mechanical Turk into the SageMaker scene for data labelling, classification and manual validation. This type of service provides basic data processing in early modelling training.
However, the Platform has also been associated with long-standing disputes, including low compensation, the quality of tasks and the governance of the Platform. Others have criticized the fact that some of the products promoted as AI are still relying on back-office manual work to complete the key steps.
Large models in turn weaken demand
In recent years, the increased capacity of large language models has complicated the role of such crowd-sourcing platforms. An analysis carried out in 2023 showed that between 33 and 46 per cent of the workers on the Mechanical Turk were using large-language models to perform their tasks, which called into question the quality of Platform data and the value of manual participation.
After the news was made public, some users stated on social platforms that the platform had long been less active, that the loss of researchers and workers and the increase in problems with robotic accounts and fraud were all the reasons for its weakness. For enterprises that rely on manual labelling processes, this also reflects that AI training infrastructure is accelerating change.