Amazon service Bezos once called 'artificial artificial intelligence' is shutting down

Amazon has announced plans to discontinue its crowdsourcing platform Mechanical Turk by the end of September 2026, according to a recent report by CNBC. Launched in 2025, the service was branded as 'artificial artificial intelligence' by its founder, Jeff Bezos, highlighting its role in outsourcing simple digital tasks to human workers.

At its peak, Mechanical Turk connected over half a million workers to various tasks, including data labeling, audio transcription, and survey completion. However, the rise of advanced AI models and the emergence of specialized data-labeling startups, such as Scale AI, Mercor, and Prolific, have led to a decline in the platform's popularity.

Jeff Bezos, the founder of Amazon, initially described Mechanical Turk as 'artificial artificial intelligence', acknowledging the platform's reliance on human labor for tasks that machines could not accomplish at the time. The name itself pays homage to the 18th-century Mechanical Turk, an 18th-century chess-playing automaton that was secretly operated by a human chess master.

The platform began as a solution to label the vast amounts of data generated by Amazon's online store. By 2026, the number of workers, known as 'trukers', had dwindled from its peak of around 500,000. Krista Pawloski, a data worker and organizer with Turkopticon, an advocacy group for data workers, noted that Mechanical Turk has been in decline due to advancements in AI and the emergence of dedicated data-labeling startups.

As the demand for AI training data grew, Amazon attempted to reposition Mechanical Turk as a data-annotation source for its SageMaker machine learning service. However, Pawloski claimed that Amazon invested fewer resources into improving the platform as competition intensified, causing many trukers to migrate to rival services. The platform's reliability has also come under scrutiny, with a 2023 study revealing that up to 46% of Mechanical Turk workers were using AI models to complete tasks that were supposed to be performed by humans, complicating the platform's role in training AI systems that use that very data.