Amazon's Mechanical Turk to End New Customer Onboarding

Amazon Web Services (AWS) announced that it will discontinue providing its crowdsourcing service "Mechanical Turk" to new customers after July 30, 2026. Details regarding how existing users will be handled have not yet been clarified at this time, but the new customer acquisition channel will be closed first.

Mechanical Turk has been known as a platform where companies can request detailed tasks from an unspecified number of people on the internet. It operates on the principle of having humans perform judgment-based tasks that are difficult for computers, such as image content verification, text classification, and voice transcription. The service name is derived from an 18th-century famous mechanical automaton that claimed to play chess. The naming was ironic, reflecting how AI was borrowing human power hidden within.

Amazon launched this service around 2005, and for nearly 20 years it has supported data collection and labeling work for researchers and companies. In particular, in the AI and machine learning fields, it has been widely utilized as a means for humans to create training data (data used to teach AI the "correct answers") for model training.

The direct reason for service discontinuation has not been publicly stated at this point. However, what can be inferred from the background is the rapid evolution of AI itself. Many tasks that previously had to be performed by humans—such as image recognition and text classification—can now be automated through large language models (LLMs) and generative AI. In other words, there is a structural shift where AI is replacing the role that Mechanical Turk once played.

Additionally, the data labeling market itself is changing, with specialized large-scale data annotation companies and more sophisticated professional services with advanced quality control emerging. These changes in the competitive environment may also have influenced the decision regarding service continuation.

Mechanical Turk can be characterized as a service emblematic of the era when AI required human labor. Its discontinuation represents a milestone indicating that the methodology of AI development itself is reaching a turning point. Going forward, the question of how to procure and manage training data for AI is expected to become an increasingly important theme across the entire industry.

Developers and researchers considering the use of this service before July 30, 2026, should realistically begin early preparations for transitioning to alternative data collection and labeling methods. How AWS provides migration support for existing users is also a matter of concern for future developments.