Breaking

Amazon to Stop Onboarding New Customers for Mechanical Turk as AI Absorbs Its Core Work

July 14, 2026 at 00:32 EDT

Sales & Business
Foundation Models
Infra & Chips

Amazon has announced it will stop accepting new customers for Amazon Mechanical Turk beginning July 30, 2026, effectively winding down a pioneering crowdsourcing platform whose defining task—supplying human labor to train AI—has increasingly been swallowed by AI itself. Existing users will not be affected and can continue to run their work, but AWS says the service is moving into a maintenance posture with no new features planned.

July 30, 2026 · Amazon Web Services

The AI-Training Platform That AI Made Obsolete

Amazon will stop accepting new customers for Mechanical Turk — the pioneering micro-work marketplace once billed as "artificial artificial intelligence." Existing users continue, but the 21-year-old service enters a slow sunset with no new features planned.

2005

Launched — one of the internet's earliest large-scale micro-work marketplaces

~21 yrs

Of powering machine-learning data pipelines before the wind-down

New features on the roadmap — maintenance posture only

AI Is Eating Its Own Feed

Automation now handles the bulk of the labeling work Mechanical Turk was built for. Model-assisted pipelines cut required human review dramatically.

100%

Human review needed
(traditional)

→

~30%

With active learning
(Ground Truth)

Up to 70% less human review — model-assisted labeling erodes Mechanical Turk's core use case.

The Irony, Compounded

Even the "human" labor was going automated. 2023 research estimated that a large share of Mechanical Turk workers were using LLMs to complete their tasks — blurring the line between human and machine labels.

⅓–½

of workers estimated to be using LLMs — raising questions about the integrity of "human" labels

Credited

Two decades of data collection and research. Institutions tapped human judgment for machine-learning work at scale — "the end of an era."

Criticized

Decline already underway — persistent bots, fraud and low-quality submissions pushed buyers to alternatives like Ground Truth, Scale AI and Labelbox.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.