Amazon Closes Mechanical Turk to New Customers July 30, Labeling the 21-Year-Old Platform a 'Service in Maintenance'

Amazon Web Services made it official earlier this week: Mechanical Turk, the crowdsourcing platform it has operated since 2005, will be closed to new customers starting July 30, 2026. AWS placed the service on its Services in Maintenance list and issued a terse statement saying it will "continue to invest in security and availability improvements" but will "not plan to introduce new features," according to TechCrunch.

In practice, this is a shutdown announcement.

What Mechanical Turk Actually Was

Launched in 2005, Mechanical Turk — named after an 18th-century "chess-playing automaton" that concealed a human player inside — was built on the same basic joke: tasks that looked automated were actually being done by people. The platform connected businesses with a distributed workforce willing to complete small, simple jobs for small, sometimes very small, payments. Solving CAPTCHA challenges. Categorizing images. Tagging sentiment in text.

Beginning in 2018, AWS repositioned it as a data annotation tool for training neural networks, marketing it alongside its SageMaker AI service, according to TechCrunch. For a period, Mechanical Turk sat at the center of how companies built early AI training pipelines.

It also had a darker footnote. The platform played a minor role in the early stages of the Facebook-Cambridge Analytica data scandal, TechCrunch reported.

The Platform Had Already Been Dying

By the time AWS made this announcement, the service was already hollowed out. A 2023 analysis found that between 33% and 46% of workers on the platform were using large language models — tools like ChatGPT — to complete the very tasks companies were hiring them for, according to TechCrunch. That discovery struck at the core value proposition of the platform: human judgment at scale.

If the workers were using AI to do the work anyway, the platform was no longer providing human annotation. It was providing a costly middle layer between businesses and AI tools those businesses could access directly.

Fraud and bots had reportedly driven away researchers and serious enterprise clients. Reddit users quoted by TechCrunch described the platform as having died "years ago." One user predicted Amazon would eventually conclude that keeping the MTurk servers running "is a waste of time and resources and pull the plug entirely."

AWS has also reportedly been closing some worker accounts without detailed explanation, according to reporting aggregated by The Register.

The Fair Concern About What Happens Next

The people most exposed by this transition are not enterprise clients. They are the workers, many of them in lower-income brackets, who have relied on Mechanical Turk for supplemental income and built reputational histories on the platform over years. AWS has not disclosed what happens to worker reputation scores, requester account balances, or pending payments when the service eventually moves past maintenance mode into a full shutdown. No final sunset date has been announced.

Worker advocates and labor researchers have long criticized Mechanical Turk for paying piece-rate wages that often fell below minimum wage by the time task selection time was factored in. A managed wind-down with unclear provisions for worker account portability or data preservation is a legitimate unresolved problem, not just a business inconvenience.

Where AWS Wants the Business to Go

The answer to "what replaces MTurk" is sitting inside Amazon's own product catalog. SageMaker Ground Truth is AWS's current data-labeling service, and it reads as the intended landing pad for existing MTurk enterprise requesters who are already inside the AWS ecosystem, according to AI Weekly. Standalone labeling vendors — Scale AI, Labelbox, and others in that space — stand to absorb the customers who want a third-party option.

Any team still quietly depending on MTurk for evaluation datasets or fine-tuning labels now has a hard deadline to migrate before maintenance mode turns into a shutdown notice.

Amazon Has Not Said Why, or When It Ends

AWS's public statement is deliberately thin. No user counts. No revenue figures. No explanation beyond "careful consideration." The company has not announced a final shutdown date, leaving existing customers in an ambiguous operational state: they can keep using the service, but on a platform the company has formally declared it will not develop further.

The unresolved question is straightforward. When AWS decides the server costs outweigh the value of keeping a dead-end service alive for its remaining customers, what notice will workers and requesters actually get?