Amazon Web Services (AWS) announced July 3, 2026, that its Mechanical Turk crowdsourcing marketplace will stop accepting new customers [1].
This move signals a significant shift in how the tech giant approaches human-in-the-loop data processing. As generative AI continues to evolve, the ability to source reliable, human-verified data has become increasingly difficult for platforms relying on crowdsourced labor.
Mechanical Turk has long served as a hub where businesses hire human workers to perform tasks that computers cannot yet do. However, the platform has faced mounting challenges regarding the authenticity of its workforce. An unnamed AWS representative said, "Not even AI can save it" [1].
The struggle to maintain a human-only workforce has been exacerbated by the very technology meant to assist it. Researchers have found that nearly half of the "turkers" on the platform were using AI to automate being human [1]. This trend undermines the core value proposition of the service, which is to provide human intelligence for tasks that AI cannot handle.
AWS did not provide a specific timeline for the complete sunsetting of the service, but the freeze on new customers suggests a managed decline. The company is now prioritizing other avenues for data labeling and human feedback, moving away from the open marketplace model that defined Mechanical Turk for years.
Existing customers may still be able to use the service, but the inability to expand the requester base limits the platform's scalability. The decision comes at a time when the industry is grappling with the quality of training data used for large language models.
“"Not even AI can save it."”
The decline of Mechanical Turk illustrates a paradox in the AI era: as the demand for high-quality human data for AI training increases, the ability to verify that a human is actually performing the work decreases. By stopping new customer intake, AWS is acknowledging that the open-market crowdsourcing model is no longer viable when workers can use AI to spoof human intelligence, necessitating a shift toward more controlled, verified data environments.
