Skip to content
NewsIncident

Amazon Closes Mechanical Turk on September 30, Ending 21 Years of Crowd-Sourced Data Labor

· by Pondero Newsdesk

The short version

Amazon confirmed it will permanently shut down Mechanical Turk, SageMaker Ground Truth, and Amazon Augmented AI on September 30, 2026. Workers and requesters have five weeks to migrate.

Amazon Closes Mechanical Turk on September 30, Ending 21 Years of Crowd-Sourced Data Labor

Amazon confirmed it will permanently shut down Mechanical Turk on September 30, 2026. The closure extends to SageMaker Ground Truth and Amazon Augmented AI on the same date, removing the full stack of Amazon's human-in-the-loop data infrastructure from AWS.

What happened

Amazon's official MTurk closure FAQ states the platform will permanently close September 30, 2026. Amazon cited a routine assessment of its programs as the reason, writing: "We regularly evaluate our programs, tools, and services and make adjustments based on those assessments." The company stopped accepting new requesters on July 30, 2026.

The September 30 deadline applies to all three products at once. Per TechTimes reporting on August 26, the closure of SageMaker Ground Truth and Amazon Augmented AI means Amazon is exiting human annotation infrastructure entirely, not just retiring one aging tool. AWS has not announced a replacement product.

For workers on the platform, the MTurk FAQ sets out the timeline: requesters can award bonuses until October 30, 2026. Workers who complete HITs before closure will receive payment on their normal disbursement schedule. Transaction histories remain accessible until January 28, 2027.

For requesters with pre-paid balances, Amazon will issue a full refund within 30 days of closure. Any HITs submitted before September 30 have until October 30 for approval or rejection; unapproved HITs auto-approve after that window.

Why it matters

Mechanical Turk underpins a significant portion of early AI history. The platform provided the labor that produced ImageNet, the image dataset that powered the deep learning era. Researcher Fei-Fei Li's team crowdsourced the labeling of 3.2 million images across 5,247 categories through MTurk, per a Quartz account of the project, producing the 2009 dataset that made the 2012 AlexNet breakthrough possible. MTurk also became a primary channel for generating the human preference data used in reinforcement learning from human feedback, the technique that made large language models coherent enough to deploy.

The closure reflects a structural shift the platform could not outlast. By 2023, researchers at EPFL estimated that 33 to 46 percent of workers on one abstract summarization task were using large language models to complete assignments and submitting the output as human judgment. The premise of the platform, that its workers provided distinctly human signal, had eroded. Amazon's decision to close Ground Truth alongside MTurk suggests the company assessed that the managed annotation business no longer justified continued investment either.

The market that absorbed MTurk's volume looks different. Per TechTimes, Mercor reached $2 billion in annualized revenue as of June 2026 by connecting frontier AI labs with vetted domain experts at rates of roughly $95 an hour for specialized reasoning tasks. MTurk served the commodity tier of that market. That tier has been largely automated.

What to watch next

Scale AI, Mercor, and Prolific are the most likely destinations for annotation workflows migrating off MTurk. Academic researchers who used MTurk for study recruitment face a harder transition: Prolific offers IRB-compatible participant pools with demographic targeting, but the cost structure differs significantly from MTurk's micro-task pricing. Watch for those platforms to announce migration tooling or onboarding incentives before September 30.

AWS teams relying on the MTurk Worker type in SageMaker labeling jobs have no in-ecosystem replacement after September 30. Amazon has not announced a successor product.

Sources