The armies of people who quietly taught artificial intelligence how to work are losing their first home. Amazon will shut Mechanical Turk, the marketplace it built in 2005 to farm out tiny digital jobs to workers around the world, on 30 September 2026. For Malaysia, which is spending heavily to grow exactly this kind of AI-data workforce, the closure is worth reading closely.
Table of Contents
What Amazon is closing
Mechanical Turk let businesses post Human Intelligence Tasks, small chores like labelling an image, transcribing a clip or answering a survey that a computer could not yet do reliably. Workers picked them up for a few cents each. Amazon founder Jeff Bezos once called the service "artificial artificial intelligence", because to the company buying the work it looked like software when it was really people. The name came from an 18th century chess-playing machine that secretly hid a human inside.
The platform has been fading for years. Amazon stopped accepting new customers in July, a move many workers read as the beginning of the end, as TechCrunch reported at the time. This is not just one aging product being retired either. Amazon's data-labelling tools SageMaker Ground Truth and Augmented AI will close on the same day, TechTimes notes, which means the company is walking away from the human-data business altogether. Coverage from Tech Startups points out that the work has shifted to specialist firms such as Scale AI, Mercor and Prolific, which recruit and vet workers for AI training rather than leaving it to an open pool.
Why this matters for Malaysia
Southeast Asia has become one of the world's busiest hubs for this behind-the-scenes AI work, and Malaysia sits near the centre of it. The country's multilingual graduates, comfortable across English, Malay, Mandarin and Tamil, are prized for annotation tasks that hinge on judgment and language, and the same workforce that powered two decades of business outsourcing has moved into labelling data for AI models. Job boards such as Jobstreet and Indeed carry a steady stream of data-annotation openings.
Malaysia is not only supplying this labour, it is investing to grow it. Oracle and MyDIGITAL have agreed to train 300,000 Malaysians in AI and cloud skills by 2029, as Fintech News Malaysia reported, with the courses folded into the Rakyat Digital platform. Budget 2026 added an RM110 million AI programme aimed at e-hailing and p-hailing gig workers, plus RM20 million in skills training, according to The Sun. The government has been open that it wants Malaysians building AI systems and not only using them, and it is handing free AI tools to young people to get there.
That is where the Mechanical Turk closure earns attention. The penny-a-task model that ran for 21 years is dying, and not because AI stopped needing people. It is because the cheap, anonymous version of the work no longer pays. The money is moving to higher-skill annotation: experts checking a model's reasoning, native speakers catching cultural errors, specialists rating answers in fields like medicine or law. For Malaysian workers, the lesson is blunt. Being inexpensive is no longer the advantage. Language, domain knowledge and reliable judgment are.
What to watch
Even the most advanced AI still leans on human feedback to stay accurate and safe, so this labour is not vanishing. It is professionalising, and the places that treat it as skilled work rather than disposable microtasks will keep the better-paid end of it. Malaysia's spending suggests it understands that. The open question is whether the training reaches the people doing the tasks today, or only the graduates arriving tomorrow.
Amazon is closing the door on AI's first human workforce. Malaysia is trying to build the next one. Whether it aims high enough will decide if those jobs are worth having.
Images courtesy of Bluestonex and phyo min on Unsplash.
