Amazon Shuts Down Mechanical Turk After 21 Years

Introduction

Amazon has confirmed that Amazon Mechanical Turk, the crowdsourced task marketplace it launched in 2005, will permanently close on September 30, 2026. The company posted the closure notice directly on mturk.com, telling both Workers and Requesters to prepare for the shutdown. The move ends roughly 21 years of a platform that quietly became foundational infrastructure for human-labeled data used to train machine learning systems, including many of today's AI models. It also lands at an odd moment: the marketplace Jeff Bezos once nicknamed "artificial artificial intelligence" is closing during the height of the actual artificial intelligence boom.

What Amazon Announced

The banner now posted on mturk.com states, in full:

"Mechanical Turk will close September 30, 2026. 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. Amazon Mechanical Turk will permanently close on September 30, 2026. For Workers and Requesters currently using the service, please visit our FAQ page to learn how you can prepare for this closure."

That is the entirety of Amazon's public explanation. The statement references a routine "assessment" of programs and services but does not attribute the decision to AI, cost, competition, or declining usage. It is a notably thin explanation for the end of a platform that operated for two decades.

The mechanics matter more than the framing. According to MTurk's official closure FAQ, Requesters can submit new HITs (Human Intelligence Tasks) only until September 30, 2026; any HITs still unsubmitted after that date automatically expire. Requesters are not charged for incomplete HITs after closure, and get 30 days post-closure to approve submitted HITs, consistent with MTurk's standing 30-day approval policy, after which unreviewed HITs are auto-approved. Requesters can still award bonuses through October 30, 2026, a full month after the platform stops accepting new work.

For Workers, the guidance is narrower: verify that a payment method and transfer schedule are set in the Worker Portal, under Account Settings > Manage Payment Preferences, because payouts will continue to follow each worker's existing schedule. MTurk says it will keep issuing tax documentation through its established processes and deadlines. All completed and approved HITs will be paid out during the normal scheduled disbursement, with no indication of an accelerated or lump-sum payout process.

For Requesters carrying a balance, Amazon says pre-paid funds will be refunded in full within 30 days, and a final bill will arrive in upcoming billing cycles. In short, this is a metered wind-down rather than an abrupt kill switch, but it places a hard deadline, September 30 for new work and October 30 for bonuses and approvals, on a service many labs, companies, and researchers have relied on for years.

Notably, this is not Amazon's first move here. AWS's own SageMaker AI features page states that Mechanical Turk, SageMaker Ground Truth, Amazon Augmented AI, and several other SageMaker features closed to new customers on July 30, 2026. That earlier cutoff drew little attention outside the SageMaker documentation. The September 30 announcement escalates that quiet boundary into a full, permanent closure of MTurk specifically. AWS's documentation does not specify a shutdown date for Ground Truth or Augmented AI beyond that new-customer cutoff, so it would be inaccurate to describe those services as closing on the same date as MTurk.

What MTurk Was and Why It Mattered to AI

Amazon describes MTurk as "a crowdsourcing marketplace that makes it easier for individuals and businesses to outsource their processes and jobs to a distributed workforce who can perform these tasks virtually." Work was broken into HITs, covering data validation, research, survey participation, content moderation, and data deduplication, among other categories, accessible through a web interface or a direct API.

The platform's importance to AI grew well beyond its original design. Amazon's own product page states that "it is easy to collect and annotate the massive amounts of data required for training machine learning (ML) models with MTurk," and highlights its role in human-in-the-loop workflows, "where human feedback is used to help validate and retrain your model," citing bounding-box annotation for computer vision datasets as an example. In practice, MTurk became a default backend for a wide range of data-labeling pipelines: image annotation, text classification, transcription, and, more recently, the kind of human-preference data used in reinforcement-learning-based model training.

Academic researchers were also heavy users. Long before it was AI infrastructure, MTurk was a workhorse of behavioral science, survey research, and social-science experiments, valued for giving researchers fast, low-cost access to a large, distributed pool of human subjects and annotators.

Amazon's own stated rationale for the service is worth noting simply because it is the company's own justification for why the marketplace existed: "While technology continues to improve, there are still many things that human beings can do much more effectively than computers, such as moderating content, performing data deduplication, or research." That rationale is being retired anyway.

Impact Analysis

For AI training-data pipelines, the closure removes a long-standing, well-documented option, though not the only one. As CNBC and Quartz have reported in covering the shutdown, MTurk's role had already been declining for years, with newer data-labeling companies, including Scale AI, Mercor, and Prolific, increasingly recruiting the workers who supply labeled data and human feedback for AI model training. That reporting frames the closure less as a sudden disruption to AI development and more as a lagging confirmation of a shift that had already occurred.

For academic and behavioral researchers, the effect may be more disruptive. Many published studies and ongoing research programs were built around MTurk-specific tooling, worker pools, and reproducibility norms. Researchers now need to migrate to alternative platforms with different worker demographics, pricing structures, and task interfaces, and re-validate that their instruments behave consistently on a new platform.

For Workers, the FAQ lays out a defined but compressed transition window. Anyone depending on MTurk income has until September 30, 2026, to complete new HITs, and needs to ensure payment preferences are correctly configured to receive final payouts. The FAQ makes no mention of severance, transition support, or a successor Amazon program for affected workers.

Pros and Cons of the Wind-Down

What Amazon has handled clearly:

• The closure notice and FAQ lay out concrete dates: no new HITs after September 30, 2026, bonuses payable through October 30, 2026.
• Existing Worker payment schedules and tax-documentation timelines continue as before, rather than changing all at once.
• Requester refunds are addressed directly, with a 30-day full-refund commitment on pre-paid balances.

What remains unclear or unresolved:

• Amazon has not stated a reason for the closure beyond a generic "assessment," despite two decades of operation.
• There is no disclosed replacement platform or transition support for Workers who relied on MTurk as an income source.
• Researchers using MTurk as academic infrastructure receive no guidance beyond the general Worker and Requester FAQ.

Outlook

The most notable part of this closure is not the mechanics, it is the timing. Coverage of the shutdown has pointed out the irony that a marketplace Bezos once called "artificial artificial intelligence" is ending just as demand for human-generated training and preference data, the raw material behind actual artificial intelligence, appears higher than ever. Whether this reflects Amazon deprioritizing this specific product line, or simply consolidating around newer AWS AI tooling, is not something the company has stated publicly. That framing should be read as industry analysis, not a confirmed motive.

What is clearer is that the data-labeling market has moved on. Specialized annotation and human-feedback companies have absorbed much of the demand MTurk originally served, and Amazon's own July 30, 2026 new-customer cutoff for MTurk, alongside SageMaker Ground Truth and Augmented AI, points to a broader retreat from this category of tooling inside AWS, even though the permanent September 30 closure is confirmed for MTurk alone.

Conclusion

Amazon Mechanical Turk's closure on September 30, 2026, ends one of the internet's longest-running crowdsourced work marketplaces, one that quietly became foundational infrastructure for machine learning data labeling and human-in-the-loop AI development. The wind-down itself is orderly, with clear dates for HIT submission, bonus payments, and refunds. What is missing is any explanation of why, and any clarity on where the researchers, Requesters, and Workers who relied on it should go next.

Editor's Verdict

Amazon Shuts Down Mechanical Turk After 21 Years earns a solid recommendation within the IT news space.

The strongest case for paying attention: clear, dated wind-down timeline: HIT submissions end September 30, 2026, bonuses payable through October 30, 2026. That alone raises the bar for what readers should expect in this space. Reinforcing that, existing Worker payment schedules and tax documentation timelines continue unchanged through closure — practical value rather than just headline appeal. The broader signal worth registering is straightforward: Amazon's closure notice cites only a routine internal assessment, without attributing the decision to AI, cost, or competition. On the other side of the ledger, one constraint is real rather than a marketing footnote: Amazon gives no stated reason for the closure beyond a generic 'assessment,' despite 21 years of operation. It should factor into any serious decision. Layered on top of that, no replacement platform or transition support is disclosed for Workers who depend on MTurk income — which narrows the set of teams for whom this is an obvious yes.

For AI industry watchers, strategy teams, and decision-makers tracking platform shifts, this is a serious evaluation candidate, not just a curiosity to bookmark. For everyone else, the safer posture is to monitor coverage and revisit once the use cases that matter to your team are demonstrated in the wild.