Amazon is winding down Mechanical Turk by closing the platform to new customers — a quiet but significant move that effectively puts the 20-year-old crowdsourced labor marketplace on a sunset trajectory. Existing requesters can continue operating for now, but no new accounts will be approved, which means teams that haven't already built workflows on MTurk can no longer start.
For AI and ML practitioners, this matters because Mechanical Turk was for years the default infrastructure for human-in-the-loop tasks: data annotation, content moderation, sentiment labeling, transcription verification, and RLHF-style preference ranking. Countless academic benchmarks and commercial training datasets were built on top of it.
The closure reflects a broader shift in how the industry sources labeled data. Specialized annotation platforms — Scale AI, Labelbox, Surge AI, Prolific, and others — have largely displaced MTurk for production-grade work, offering better quality controls, worker vetting, and audit trails that enterprise AI teams require. MTurk's low cost came with well-documented quality variance that made it increasingly hard to justify for high-stakes labeling jobs.
If your team still relies on MTurk for any part of a data pipeline, now is the time to migrate. Evaluate alternatives based on your specific task type: Prolific is strong for research and survey-style tasks; Scale and Labelbox suit structured annotation at volume; Surge AI positions itself for higher-complexity NLP work. Factor in cost-per-label, worker agreement rates, and how each platform handles edge-case quality control.
The deeper takeaway is that the era of cheap, anonymous crowdsourcing as a foundation for AI training data is effectively over. Regulatory pressure around data provenance, model transparency requirements, and hard lessons from low-quality training sets have pushed the industry toward more accountable, auditable labeling workflows — and MTurk's architecture was never built for that world.