SEATTLE — Amazon has officially announced that it will permanently shutter Amazon Mechanical Turk (MTurk), the pioneering crowdsourcing marketplace that fundamentally altered how developers, researchers, and enterprises approach micro-tasks and human-in-the-loop computing. According to a closure notice posted to the platform’s official site, the service will officially cease operations on September 30, 2026.
The impending shutdown marks the end of an extraordinary run for a platform that introduced the tech industry to the concept of "Artificial Artificial Intelligence"—a tongue-in-cheek term coined by Amazon to describe automated software workflows that ultimately relied on human cognition behind the scenes. For nearly two decades, MTurk served as an invisible engine for the digital economy, matching a distributed global workforce of independent contractors with organizations needing help with data annotation, transcription, content moderation, and academic survey research.
However, shifting industry demands, the rapid evolution of generative artificial intelligence, and corporate re-evaluations have rendered the platform obsolete in Amazon’s eyes. As the tech giant prepares to pull the plug, millions of workers and thousands of requesters must navigate a structured, multi-month wind-down process that spans financial settlements, API deprecations, and the retirement of integrated machine learning pipelines.
Executive Overview
The decision to shutter MTurk was reached following routine internal portfolio reviews. In a succinct statement published on the platform, Amazon noted: "We regularly evaluate our programs, tools, and services and make adjustments based on those assessments. Following an assessment, we’ve made the decision to close Amazon Mechanical Turk, effective September 30, 2026."
The shutdown will unfold in carefully managed phases over the remainder of 2026 and into early 2027. While September 30, 2026, marks the hard deadline for human intelligence task (HIT) submissions, downstream administrative processes—such as work approvals, bonus disbursements, and prepaid account refunds—will stretch through October 2026. Historical transaction data will remain accessible to users until January 28, 2027.
Crucially, the closure is not limited to the standalone MTurk web portal and API marketplace. It extends deeply into Amazon Web Services (AWS) infrastructure. Specifically, the Amazon Mechanical Turk Worker type will be permanently retired from Amazon SageMaker Ground Truth and Amazon Augmented AI (A2I). Organizations that built automated machine learning data labeling pipelines on top of these AWS services will be forced to migrate to alternative workforce sources before the September deadline.
Detailed Chronology and Shutdown Mechanics
To ensure an orderly departure from the market, Amazon has released a comprehensive FAQ outlining the operational wind-down for both sides of the marketplace: Workers and Requesters.
The Worker Wind-Down
For the distributed workforce—often referred to colloquially as "Turkers"—the primary objective before the closure is verifying payment details. Amazon has instructed workers to log into the MTurk Worker Portal to confirm their preferred payment method, transfer schedule, and payout frequencies.
- Final Submissions: Workers can continue completing and submitting HITs up until September 30, 2026. Any unsubmitted or draft HITs remaining in queues after this timestamp will automatically expire.
- Disbursements: All completed and approved HITs will be paid out according to each worker’s normal scheduled disbursement cycle. Tax-related documentation will continue to be distributed by MTurk through established processes and statutory deadlines.
The Requester Wind-Down
Organizations and developers utilizing MTurk to source human labor face a tight timeline to wrap up projects, settle accounts, and extract historical data.
- Work Approvals and Bonuses: Requesters retain standard policy rights to approve or reject completed work for 30 days following the submission cutoff, extending through October 30, 2026. If a requester takes no action during this window, pending HITs will be automatically approved. Requesters can also issue performance bonuses to workers up until this October 30 deadline.
- Financial Settlements: Requesters must verify their billing and payment information within the MTurk Requester Portal. Pre-paid requester balances will be refunded in full within 30 days of closure. Any outstanding balances or final transactional fees will appear on upcoming AWS billing cycles.
- Data Retention: Transaction histories, export logs, and administrative records will remain accessible to requesters through January 28, 2027, after which all platform data will be scrubbed.
Supporting Context & Metrics: The Mechanics of "Artificial Artificial Intelligence"
To understand the cultural and technological weight of MTurk’s departure, one must look back at its origins and economic architecture. The service derived its name from an 18th-century hoax: in 1769, Hungarian nobleman Wolfgang von Kempelen constructed a mechanical chess-playing automaton that traveled across Europe, defeating formidable opponents including Benjamin Franklin and Napoleon Bonaparte. Kempelen frequently opened the cabinet doors to display an intricate maze of gears, cogs, and springs, convincing spectators they were witnessing true machine intelligence. In reality, a master chess player was skillfully concealed inside.
Amazon adapted this historical metaphor for the digital age. Launched in 2005, MTurk was originally conceived to solve internal data-processing problems at Amazon that required human judgment—tasks that remained frustratingly difficult for computers. While algorithms could easily process petabytes of structured data, they struggled immensely with tasks that children master before learning to speak, such as identifying objects in photographs, transcribing low-quality audio, or detecting nuanced hate speech in forum posts.
The Platform Model and Pricing Architecture
MTurk operated on a simple yet powerful transactional model. Requesters—ranging from academic researchers at Ivy League universities to Fortune 500 tech companies training computer vision models—submitted tasks via a web interface or a RESTful API. To a developer writing code, submitting a task looked identical to making a remote procedure call (RPC). Behind the curtain, however, a human worker completed the assignment in exchange for a cash reward.
The platform’s pricing structure was divided into two main components:
- The Worker Reward: An amount independently set by the requester to incentivize completion.
- The Platform Fee: A baseline commission paid to Amazon, typically standard at 20% of the total reward and bonus amount.
Additional fees applied for scale and quality control:
- A surcharge of an extra 20% on the reward was levied for HITs carrying 10 or more assignments.
- A minimum fee of $0.01 was enforced per assignment or bonus payment.
- Specialized quality filters, such as the Masters Qualification—awarded to workers who maintained exceptionally high accuracy across diverse tasks under ongoing statistical monitoring—added an extra 5% fee to the reward.
- Premium Qualifications incurred an additional per-assignment fee starting at $0.05.
Despite these costs, MTurk offered an unmatched global, on-demand, 24/7 workforce. Its use cases spanned machine learning dataset annotation (such as drawing bounding boxes for autonomous vehicle computer vision models), business process outsourcing, data validation, market research surveys, and extensive web content moderation.
Integration and Fallout for AWS Machine Learning Pipelines
One of the most consequential aspects of the September 2026 shutdown is the severing of MTurk’s integration with enterprise-grade AWS machine learning toolchains.
For years, Amazon SageMaker Ground Truth and Amazon Augmented AI (A2I) relied on the MTurk workforce as a core option for building human-in-the-loop data pipelines. When machine learning models encountered ambiguous data points during training or inference—such as a self-driving car’s computer vision algorithm failing to categorize an obstructed road sign—A2I workflows could automatically route those edge cases to human reviewers via the MTurk network.
As of September 30, 2026, the "Amazon Mechanical Turk Worker type" option will be permanently deprecated within SageMaker Ground Truth and A2I creation workflows. Engineering teams that built automated data labeling and review loops directly upon this infrastructure will have to scramble to re-architect their pipelines. Enterprises will be forced to transition either to private, internal workforce pools, alternative third-party data annotation vendors (such as Scale AI, Labelbox, or Appen), or entirely automated synthetic data generation models.
Official Statements and Industry Reactions
Reaction across the tech, academic, and developer communities has been a mixture of nostalgia and pragmatic resignation. For academic researchers who relied on MTurk for behavioral and psychological studies, the platform’s closure represents a significant disruption. For years, MTurk was praised in academic literature as a cost-effective, diverse participant recruitment pool, though it faced mounting criticism in later years regarding data quality, bot infiltration, and fair compensation.
Labor advocates have long criticized MTurk for its lack of traditional labor protections. Because workers were classified globally as independent contractors rather than employees, they lacked access to a guaranteed minimum wage, health insurance, sick leave, or protections against arbitrary rejections and unpaid work by dishonest requesters. While platforms like Turkopticon emerged to help workers rate requesters and protect themselves against wage theft, the fundamental precarity of "microwork" remained a persistent ethical flashpoint for the platform.
Industry analysts note that MTurk’s closure is emblematic of a broader technological shift. The rise of Large Language Models (LLMs) and advanced multimodal generative AI systems has dramatically reduced the market demand for traditional human micro-tasking. Tasks that once required armies of human workers to manually tag thousands of images or moderate basic text strings can now often be handled, validated, or pre-filtered by automated foundation models at a fraction of the cost and speed.
Future Outlook: The Post-MTurk Landscape
As the countdown to September 30, 2026, continues, the digital economy stands on the precipice of a post-MTurk reality.
For requesters, the immediate priority is auditing existing AWS workflows, backing up historical transaction data, and transitioning data annotation pipelines to modern SaaS labeling platforms. For the displaced global workforce, the closure forces an uncertain transition toward alternative gig economy platforms, specialized data-labeling startups, or emerging decentralized AI training networks.
Amazon Mechanical Turk began as an audacious experiment in blurring the lines between human cognition and computational machinery. Two decades later, as artificial intelligence grows increasingly autonomous, the original pioneer of "Artificial Artificial Intelligence" has quietly reached the end of its useful lifecycle—leaving behind a transformed technological landscape that it helped create.
Final Timeline at a Glance
- September 30, 2026: Official platform closure. Final day for HIT submissions and removal of MTurk worker types from SageMaker Ground Truth and Amazon A2I.
- October 30, 2026: Final deadline for requesters to approve or reject completed work (auto-approval kicks in thereafter) and issue worker bonuses.
- Within 30 Days of Closure: Full refunds processed for prepaid requester balances.
- January 28, 2027: Transaction history and historical portal data permanently taken offline.
