Amazon shuts down crowdsourcing platform Mechanical Turk after 21 years

Background of the platform

Mechanical Turk was one of the first large-scale platforms through which companies and researchers could purchase human labor for so-called microtasks. These included annotating images, categorizing texts, or validating search results.

The name of the platform references a historical curiosity: Allegedly fully automatic, the device — in the form of a puppet dressed in Ottoman garments — played chess against people around the world from 1770 onward.

MTurk came under increasing discussion in recent years because the platform was also used to create and process AI training data. Studies and research had previously documented what were at times precarious working conditions and low hourly wages for crowd workers.

Significance for AI training data

With the shutdown, a pioneer of the crowdsourcing model loses its operational base. Industry observers see this as an indication of the shift in the procurement of data for AI models, which is increasingly being covered by specialized providers or synthetic data sources.

For active clients and workers, the end of MTurk means switching to alternatives. Platforms such as Appen, Toloka, or Clickworker are mentioned in reporting. The exact timeline for the shutdown as well as possible data migrations are not detailed in the available sources.

Consequences for clients and workers

Amazon itself does not substantively comment on the discontinuation of MTurk in the available reports. Observers see the move as a logical step, since most former MTurk tasks are now covered by internal pipelines or specialized service providers.

In recent years, Amazon had already scaled back individual areas of the platform and restricted new features. The now-announced complete shutdown thus follows a longer trend of reducing external access.

The closure of MTurk is seen as a symbol for the end of an era in which human micro-labor on the internet was a mass phenomenon. It remains open whether the remaining order volume can be fully absorbed by specialized data platforms.