Amazon closes Mechanical Turk to new customers
Amazon’s decision to stop onboarding new Mechanical Turk customers marks a shift away from the old crowdsourced data-labelling model that helped train earlier generations of machine learning systems.
One of the internet’s earliest human-in-the-loop work platforms is entering managed decline just as AI is reshaping the labour model it helped create.
What happened
Amazon will stop accepting new Mechanical Turk customers on July 30, 2026. Existing customers can continue using the service, but AWS says it does not plan to introduce new features.
Mechanical Turk launched in 2005 as a marketplace for small tasks that computers struggled to perform. Over time, those tasks included surveys, classification and data annotation, making the platform part of the wider machine-learning data economy.
Why it matters
This is a useful AI-market signal because the data stack has changed significantly since Mechanical Turk became popular.
Model developers increasingly need more specialised annotation, expert feedback, domain-specific evaluation and higher-quality human review. At the same time, synthetic data and model-assisted labelling are reducing the value of some low-complexity crowd tasks.
The platform also came to represent a trust problem inside the data pipeline. As generative AI tools became widely available, it became harder to know whether supposedly human-completed work had actually been produced with AI assistance.
The bigger picture
AI is changing the economics of human data work. The market is moving away from large pools of anonymous microtask workers toward more specialised contributors, expert evaluators and tightly managed feedback systems.
Mechanical Turk’s retreat from new customers does not mean human input is becoming less important. It suggests the opposite: as models become more capable, the remaining human work is becoming more specialised, more quality-sensitive and harder to commoditise.