- Closure confirmed. Amazon will shut down Mechanical Turk on September 30, 2026, ending a platform that has been the invisible backbone of machine-learning dataset construction for two decades.
- AI replacing AI’s own teachers. The tasks MTurk once routed to human workers — data labeling, audio transcription, image classification — are now handled cheaply enough by AI systems that the economics of the platform no longer hold.
- Wider sweep. Amazon is closing MTurk alongside two related services: SageMaker Ground Truth and Amazon Augmented AI, suggesting a deliberate exit from the human-in-the-loop data-annotation market.
Amazon announced on Monday that it will close Mechanical Turk on September 30, 2026, citing an internal assessment of its programmes, tools, and services. New customer sign-ups had already been halted on July 30 alongside the concurrent closures of SageMaker Ground Truth and Amazon Augmented AI, CNBC reported. The platform’s 21-year run makes it one of the longest-lived products in Amazon Web Services’ history, and its shutdown marks the clearest signal yet that the crowd-sourced data-labeling model has been structurally displaced by the tools it helped create.
What Mechanical Turk Was
Jeff Bezos launched Mechanical Turk in November 2005, describing it as “artificial artificial intelligence” — a wry acknowledgment that the platform was routing to human workers the tasks that computers could not yet handle on their own. Workers, known as Turkers, completed what Amazon called Human Intelligence Tasks: labeling images, transcribing audio and video, verifying addresses, answering surveys. Individual task payouts typically ranged from a fraction of a cent to a few dollars, and at the platform’s peak the workforce was estimated at several hundred thousand people globally.
The service became foundational to the machine-learning industry. Annotated datasets produced on MTurk underpinned training runs at universities and AI companies across the world, including some of the models that would eventually automate the very tasks the platform assigned.
The Economics That Killed It
The shutdown reflects a structural shift rather than a commercial failure. Over the past three years, large language models and multimodal AI systems have become capable enough to handle most of the tasks MTurk was used for — labeling, classification, transcription — at marginal costs well below what the platform’s workforce charged. The concurrent closure of SageMaker Ground Truth, which allowed companies to build custom labeling workflows combining human and machine annotation, confirms that Amazon has concluded the human-in-the-loop annotation market no longer justifies dedicated infrastructure.
Amazon’s statement was terse: “Following an assessment, Amazon made the decision to close AWS Mechanical Turk, effective September 30, 2026. We regularly evaluate our programs, tools, and services and make adjustments based on those assessments.”
What Workers and Requesters Face
Turkers still active on the platform have until September 30 to complete pending tasks and withdraw their earnings. For requesters — the companies and researchers that published tasks — the closure creates a migration challenge, particularly for academic projects that relied on MTurk’s scale and pre-vetted worker pool. Several third-party annotation platforms have moved quickly to court displaced customers, but none replicates the platform’s combination of size, integration with AWS workflows, and the decade-long behavioral data that requesters used to assess worker quality.
The closure completes a long decline for a service that once defined the frontier of human-computer collaboration. That frontier has moved: the tasks that required human judgment in 2005 are today handled by foundation models, and the tasks that require human judgment in 2026 are considerably harder to mechanise. Whether a platform emerges to match human workers to those harder problems — or whether AI continues to push the boundary outward — will shape the next phase of what Bezos once called the “artificial artificial intelligence” market.