Amazon closes Mechanical Turk to new clients: end of the crowdsourcing AI era

Amazon announced on July 5, 2026 that it is discontinuing the recruitment of new clients for Mechanical Turk—a crowdsourcing service that for over twenty years has supplied the AI industry with human data labeling. TechCrunch, which reported the decision, called this moment "possibly the final days" of the platform.

What is Mechanical Turk

The MTurk platform was launched by Amazon in 2005. The name refers to "The Turk"—an 18th-century mechanical chess automaton that appeared to be a machine but was actually operated by a hidden human inside. The principle of MTurk was similar: outwardly a technological service, inwardly a global network of workers performing micro-tasks for modest compensation.

Companies posted tasks on the platform that algorithms at the time could not solve reliably:

• labeling images, audio, and texts
• transcribing video and audio recordings
• moderating user-generated content
• evaluating search result quality
• collecting feedback for language model training (RLHF)

MTurk provided a significant portion of the data on which early language models were built. Human ratings through crowdsourcing became the foundation of the RLHF method—the approach companies used to align AI system behavior with real people's expectations. Many modern language models went through similar labeling processes in their early development stages.

Why now?

MTurk's peak came during the period when AI systems required vast amounts of manually created data, and only humans could handle this with sufficient quality. Today the situation has fundamentally changed. Major labs are transitioning to synthetic data: powerful models themselves generate training examples, evaluate their quality, and fine-tune the next generation of systems with minimal human involvement.

Crowdsourcing labeling is turning into a niche tool rather than the backbone of the industry. At the same time, more specialized platforms have emerged—with better quality control and professional annotators focused on complex tasks. MTurk, designed for mass flows of cheap tasks, found itself disadvantaged on both sides of the market.

Platform workers were called "Turkers." At their peak, there were hundreds of thousands worldwide—from the USA to India and Bangladesh. For many it was supplementary income; for some it was the primary source. Halting new client recruitment means a gradual depletion of new tasks on the platform.

What this means

Mechanical Turk embodied an entire era in AI development: technologies relied on massive distributed human labor. Amazon's decision to halt new client recruitment confirms what has been observed for several years: the industry is moving away from crowdsourced labeling toward synthetic data and self-learning systems.

Amazon has not announced a complete MTurk shutdown—existing clients will likely be able to continue operations. But stopping new recruitment means a slow decline rather than an abrupt end. Symbolically: the platform that helped create modern AI is being displaced by AI itself.