Amazon Mechanical Turk closed on 2026-09-30. Here is how to keep your work running →

Guide

MTurk closed today: how to keep your Ground Truth and A2I workflows running

Deepen AI · Published 2026-09-30

Facts checked against Amazon Mechanical Turk and AWS notices on 2026-09-30.

Amazon Mechanical Turk closed permanently today, 2026-09-30. It stopped accepting new requesters on 2026-07-30. For many teams, the harder question is not MTurk itself. It is the human step inside SageMaker Ground Truth labeling jobs and Amazon Augmented AI (A2I) review loops that relied on the MTurk workforce.

This guide covers what changed, what did not, and every realistic option for keeping that work running. It is written for teams with business and machine learning tasks. If you ran academic surveys, skip to the last section.

The dates that still matter

  • 2026-09-30: MTurk closed permanently.
  • Until 2026-10-30: you can still approve or reject completed HITs. Do it now, so the people who did the work get paid and your records are clean.
  • Until 2027-01-28: your transaction history is still available. Download it for finance and audit records.

What changed in Ground Truth and A2I, and what did not

Ground Truth and A2I are not closing. What is gone is one workforce option: the Amazon Mechanical Turk workforce, which sent your tasks to the public MTurk pool. Labeling jobs and human review workflows that used it now need a different workforce.

AWS supports two other workforce types in the same services:

  • Private workforce. People you choose, who sign in to a labeling portal for your account. You manage them with Amazon Cognito or your own OIDC identity provider.
  • Vendor workforce. A labeling vendor you subscribe to through AWS Marketplace.

Your task templates, input and output buckets, and the rest of your pipeline stay the same. The workforce setting is what changes.

Step 1: find everything that used the MTurk workforce

Search for the public workforce ARN. It ends in workteam/public-crowd/default. Check three places:

  1. Labeling job definitions, and the scripts or notebooks that create them.
  2. A2I human review workflows (flow definitions).
  3. Infrastructure code: CloudFormation, CDK or Terraform.

Also search for the public workforce task price setting. It only applied to MTurk.

For each hit, write down the owner, weekly volume, task type, and whether the data includes personal information. You need all four to choose a workforce.

Step 2: choose a new workforce

There is no single right answer. It depends on volume, data sensitivity, and who you want accountable for quality.

Option A: your own staff as a private workforce. Best when the task needs deep product knowledge, or the data must stay with a small group. You get full control. You also take on hiring, training, scheduling and quality review.

Option B: a managed team as a private workforce. A vendor provides people who sign in to your portal as members of your private workforce. You keep control of access. The vendor handles training, a team lead and QA. Ask any vendor how they measure quality, whether the same people stay on your work, and what their security audits cover.

Option C: a vendor workforce from AWS Marketplace. Useful if you want to buy through your AWS account. Check which task types each vendor supports and how pricing works before you commit.

Option D: automate more, review less. If a model is now good enough for most items, send only low-confidence items to people. This is what A2I was built for. You still need a workforce for the remainder, so this combines with A, B or C.

Option E: move the task out of AWS labeling tools. If Ground Truth was only there to reach MTurk, it may be simpler to have a team work directly in your own tools, such as a spreadsheet, an admin panel or a labeling tool you already run. You give up the AWS job orchestration in exchange.

Step 3: set up the private workforce

In the SageMaker console, open Labeling workforces and choose Private. If you do not have a private workforce yet, create one with Cognito or your OIDC provider. Each AWS account has one private workforce per Region, and it can hold several work teams.

Create a work team for the new people and invite them by work email. Each person gets an invitation to your labeling portal. If your security policy requires it, limit portal sign-in to set IP address ranges.

Step 4: point your jobs and workflows at the new team

  • Ground Truth: create new labeling jobs with the private work team ARN. Leave out the public workforce task price.
  • A2I: an existing flow definition cannot be edited. Create a new human review workflow that uses the private work team ARN and the same task template (human task UI). Then update your code to call the new workflow ARN.

Check any job that was still running on the MTurk workforce at closure, and decide whether to rerun its unfinished items.

Step 5: test before you move everything

Send a small batch first. Build a gold set: 50 to 200 items with known-correct answers, checked by someone who knows the task. Mix gold items into the test batch, compare results, and fix the instructions wherever people disagree. Then move the rest of the volume.

Keep the gold set after the move. It is the simplest way to see whether quality drifts later, whoever does the work.

What you give up, and what you gain

MTurk offered a very large pool on demand and per-task pricing. No private workforce matches that for short, spiky bursts. What you gain is a known group of people who learn your rules, someone accountable for quality, and a clear record of who did each task.

If you ran academic surveys

Participant recruitment for research is a different job. Prolific and CloudResearch are built for it, with participant pools and screening tools.

A one-week checklist

  • ·Approve or reject completed HITs (deadline 2026-10-30)
  • ·Download your MTurk transaction history (deadline 2027-01-28)
  • ·List every job, workflow and infrastructure file that references public-crowd/default
  • ·Choose a workforce option for each task
  • ·Create or update your private workforce and work teams
  • ·Recreate A2I human review workflows with the new work team
  • ·Build a gold set and run a test batch
  • ·Move the remaining volume

Disclosure: we run Deepen AI Services, a managed team that joins customer portals as a private workforce for business and data tasks (option B above).