Uber tests AI training tasks for drivers in new pilot program

Uber has launched a pilot program in the US allowing drivers and couriers to earn money by completing digital tasks in the Uber Driver app.

These tasks include uploading photos, recording short voice clips, and submitting documents to help train AI models.

Uber, which started testing the program in India, plans to expand it in the US by the end of 2025.

The company said each task pays a fixed amount based on complexity and time required, with payments appearing in drivers’ balances within 24 hours.

Task availability will depend on client demand for data labeling jobs, and pay rates for US drivers have not been finalized.

The program is not tied to Uber’s autonomous ride partnerships, such as its collaboration with Waymo.

Uber’s AI Solutions platform provides labeled data to clients including self-driving tech firms Aurora and Tier IV.

Data labeling is a growing market, valued at nearly US$4 billion in 2024 and expected to reach over US$17 billion by 2030.

🔗 Source: Axios

🧠 Food for thought

Implications, context, and why it matters.

Uber says availability depends on “client needs”, so work can be sporadic 1. Each job is optional, with time and pay estimates shown upfront 1.
The setup looks like microtask hubs such as Amazon Mechanical Turk, which pays cents to a few dollars for small online jobs. For most people this is side money, not a main paycheck.
Uber has not shared throughput rates or guaranteed task volume, so drivers and couriers face income swings. Hourly take often hinges on how many jobs appear and how fast workers accept them.

Uber AI Solutions is the company’s data services platform 2. It offers labeling in 100+ languages and access to more than 8 million “experts” 2. In practice these are drivers with couriers doing annotation work, which helps AI startups and enterprise groups gather culturally specific datasets. Builders of multilingual large language models (LLMs) or regional AI products can use this distributed workforce for localized audio, image, and text collection.
The menu lists 30+ capabilities 2. Examples include sentiment detection plus Light Detection and Ranging (LiDAR) annotation. You also get synthetic data generation (computer-generated training data) 2. Many buyers may view it as a Scale AI alternative for custom datasets.
Companies can contact Uber AI Solutions to discuss service-level agreements (SLAs), programmatic task routing via APIs, and pricing for high-volume annotation projects 3.