AI firms are hiring thousands of gig workers to train robots using wearable tech

Somewhere in Tamil Nadu, India, a contract worker is wearing a head-mounted camera while folding laundry. Not for a quirky YouTube channel. For a robotics company that needs to teach a humanoid machine how human hands actually work.

AI companies are building an entirely new layer of the gig economy, one where the product isn’t code reviews or content moderation but raw footage of human beings doing ordinary things. Washing dishes. Assembling parts. Opening jars. The mundane choreography of daily life, captured frame by frame, is becoming one of the most sought-after training resources in robotics.

The new data gold rush

Companies like Micro1 and Objectways have recruited thousands of workers across more than 50 countries to record their hand movements and physical tasks using wearable cameras and motion sensors. The footage feeds directly into training pipelines for humanoid robots, giving machines the visual and spatial data they need to replicate complex manipulation tasks.

Tesla, Figure AI, and Scale AI are among the firms purchasing these real-world datasets. Robotics companies collectively spend more than $100 million annually on this kind of training data, a figure that reflects just how difficult it is to teach robots physical dexterity through simulation alone.

The sector’s appetite for this data tracks with the broader investment surge in humanoid robotics. Over $6 billion flowed into the humanoid robotics space in 2025, a number that helps explain why companies are willing to pay serious money for videos of people doing chores.

A global workforce earning pennies on the dollar

In India, contract workers earn roughly 250 rupees per hour, which comes out to about $2.60. That’s the going rate for strapping on a sensor-laden headset and recording yourself performing household or factory tasks for hours at a time.

Objectways has established data collection operations in regions like Tamil Nadu, scaling up the volume of footage that can be gathered at relatively low cost. Micro1 operates across more than 50 countries, tapping into labor markets where the pay, while modest by Western standards, still attracts workers looking for flexible income.

The pattern echoes what happened with content moderation and earlier waves of AI data labeling. Companies in Silicon Valley and Shenzhen need vast quantities of human-generated data, and they find it most cost-effective to source that labor from developing economies.

Privacy concerns are already surfacing

When the workplace is your living room and the recording device is mounted on your head, privacy gets complicated fast. Workers have raised concerns about what happens to footage captured in their personal environments. A camera recording hand movements while cooking dinner is also, inevitably, recording the kitchen, the family members walking through, and whatever else falls within its field of view.

Questions about data ownership, consent, and downstream usage remain largely unanswered. Workers often don’t know exactly which companies will ultimately use their footage, how long it will be retained, or whether it could be repurposed for applications beyond robotics training.

The race for embodied AI data is going global

This isn’t purely a Western phenomenon. Researchers in China have been using VR headsets and exoskeletons to gather similar movement data, pursuing the same goal through slightly different hardware.

The $100 million-plus that robotics firms spend annually on these datasets is likely to grow as more companies enter the humanoid robot race and existing players push toward commercial deployment.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.