Amazon Is Closing Mechanical Turk to New Users — and It's About AI

Amazon has announced it will stop accepting new customers for Mechanical Turk, its long-running crowdsourced labour platform. If you know what Mechanical Turk is, this probably doesn't surprise you. If you don't, the story of why Amazon is winding it down tells you something important about where AI has arrived.

Mechanical Turk, launched in 2005, was built on a quietly radical idea: that there are certain tasks computers can't do well, but humans can do quickly and cheaply at scale. Amazon built a marketplace where businesses could post these tasks, things like labelling images, transcribing audio, categorising text, verifying addresses, and pay a distributed global workforce to complete them for small per-task fees. The workers, known as Turkers, were largely invisible. The platform was named after an 18th-century chess-playing automaton that concealed a human operator inside.

The name aged better than anyone probably expected.

What Mechanical Turk Was Actually Used For

For a long time, Mechanical Turk was the backbone of AI training data collection. The irony was always somewhat uncomfortable: the machine learning models that were supposed to make human labour obsolete were being trained on data labelled by underpaid human workers on Mechanical Turk. Researchers at universities used it to run experiments. Companies used it to build datasets. Content moderation teams used it to classify disturbing material that automated systems couldn't reliably handle.

At its peak, Mechanical Turk had hundreds of thousands of active workers globally. Research consistently found that many earned well below minimum wage once you accounted for the time spent finding and qualifying for tasks. The platform attracted significant criticism for its labour practices, and there were periodic calls for Amazon to introduce minimum pay standards or basic worker protections. Most of those calls went unanswered.

Why It's Closing to New Users Now

Amazon hasn't provided a detailed explanation, but the timing and context make the picture fairly clear. Generative AI systems, particularly large language models and multimodal models, have become capable enough to handle most of the tasks Mechanical Turk was built to facilitate. Image labelling, text classification, transcription, sentiment analysis: all of these are now handled faster, cheaper, and more consistently by AI tools than by human microworkers.

The market for human data labelling hasn't disappeared entirely. Highly specialised tasks, particularly those requiring cultural nuance, domain expertise, or the kind of edge-case judgement that models still struggle with, continue to employ human labellers. But those tasks increasingly go to more structured data annotation companies like Scale AI or Appen rather than to open crowdsourcing platforms.

Mechanical Turk closing to new users doesn't mean it shuts down overnight. Existing workers and clients will presumably continue using the platform until Amazon makes further announcements. But the message is clear: the era it was built for is ending.

The Human Cost

It would be easy to treat this as a tidy technology story about AI replacing an older workflow. It's also a labour story. The workers who relied on Mechanical Turk earnings, many of them in developing countries where even small per-task payments are meaningful, are facing the loss of a platform without much in the way of transition support or safety net.

This is a pattern that's played out across multiple sectors as AI has improved: the labour most vulnerable to displacement tends to be informal, low-paid, and held by people with the fewest alternatives. Mechanical Turk workers don't have unions, don't have contracts, and don't have the visibility that would make their displacement a political issue in the same way that, say, automotive workers do.

What This Signals More Broadly

The closure of Mechanical Turk to new users is a small but concrete indicator of AI's real-world economic impact. It's not a hypothetical about future disruption. It's a platform that was central to AI's development, built on human labour, and is now being made redundant by the very systems it helped create. The machine has learned. The humans inside it are being let go.