Mohammed Ashim is 22 and has been cutting hair and trimming beards in Bengaluru for four years. Recently he started strapping a camera to his head while he works.
The camera records only his hands and fingers. His clients’ faces are blurred out. He wears it five or six hours a day.
“The routine is simple, but you have to be disciplined about how you move,” he told the BBC’s Priti Gupta. “The operations team taught me that my hands always have to stay cleanly inside the frame, the lighting has to be sharp, and I can’t tilt my head too abruptly while working.”
“It’s good money, but honestly, the feeling of doing something this futuristic is what keeps me excited. I never imagined my razor techniques would be useful to a global tech company.”
What the footage is for
Ashim is one of hundreds of thousands of people worldwide now paid to record their own work. The video is collected by local companies, processed, and sold on to technology firms and research labs training physical AI — the kind that controls robots in the real world.
The reason this market exists is a gap in the data. Text AI was trained on the internet. There is no internet for this.
“There is no existing internet archive of embodied human activity — no dataset of a kitchen being cleaned, or a car being assembled by hands that know exactly how much force to apply,” said Manish Agarwal, founder of Humyn Labs, which has bases in Bangalore and San Francisco. “That data has to be captured from scratch, in the real world, across diverse environments.”
And simulation is not a substitute. “Physical tasks require high-quality signals involving limb movements, fingers, wrists, elbows, shoulders and dynamic force application,” Agarwal said. “For example: applying the correct grip force so a plastic bottle doesn’t crumble versus a steel glass.” Sometimes collectors wear hardware on their hands and fingers to capture contact itself.
The rates
Rashdan Technologies pays $5 to $10 for eight to ten hours of recorded work. “For a lot of these people, that’s meaningful,” said co-founder and chief executive Shahil Baig.
Humyn Labs has 96,000 data collectors across Latin America and South East Asia. Agarwal expects dozens more firms to enter the market in India this year.
The buyers are specific about what they want. “Every robotics company, every frontier AI lab, has a specific kind of data they want: industrial tasks, agriculture, household chores, small business work like pottery, garment-making, embroidery, candle-making. We go wherever the task is,” Baig said.
Geography is part of the specification. “You need to source across a variety of factories, a variety of tasks, in different geographies because if you want to build deployment for US factories or US houses, you’d want environments that are a proxy for that,” Agarwal said. “Residential houses in India may not be the right context for building a robotics use case for the US.”
Consent, and the direction it is moving
Baig says consent matters and describes the safeguards: “We record hands, not faces, and where multi-camera rigs do capture more of the body or environment, we blur faces out. No names, no contact details, nothing that identifies them personally goes into what we hand off.”
But the ask is growing. “Now clients are asking for four to six cameras rigged around a person’s head to cover the entire scene, not just the hands,” he said. “Getting someone to agree to one small camera is one conversation. Getting them to agree to wearing what’s effectively a five-camera headset is a completely different, and harder, one. We pay more for that tier of data collection because of what we’re asking of people.”
The question nobody can answer yet
Are these workers training the machines that will replace them?
Bino Paul, senior professor at the School of Management and Labour Studies at the Tata Institute of Social Sciences in Mumbai, says it is too early to tell. AI may be a useful tool for some workers and may automate other jobs outright. His worry is about timing rather than outcome.
“By the time workers realise what they have participated in, things may have gone too far,” he said. He wants governments to regulate: “AI cannot be viewed merely as a technical system. We need to develop a socio-technical system in which ethics is upheld through human agency.”
Mohammed Arman, who has worked in textiles in Bangalore for ten years, has already done the calculation himself. “Garments stitching and cutting. I know this work so well I could probably do it with my eyes closed,” he said. He now wears a head-mounted camera while he does it.
“For me, it’s simple. This is survival. I have a family back home in Bihar, and every rupee I send back matters. If someone offers me extra money to wear a camera while I do the exact same work I’ve been doing for 10 years, I’m not going to say no. It doesn’t cost me anything.”
Why Bangladesh should read this closely
Note which trades the buyers named: garment-making and embroidery. Bangladesh has roughly four million people doing precisely that work, at a scale India does not match, which makes this country an obvious next source of exactly the footage being bought.
So there is a decision to be taken before the offers arrive, not after. What are the terms: who owns the recording, what it may be used for, whether a factory can make participation a condition of a shift, and what the rate should be when the buyer is a frontier lab and the seller is on a sewing line. A worker signing away rights for $5 is not a scandal in itself. A sector doing it without having discussed it is a missed chance to set the price.
Related: Bangladesh’s draft AI policy and the garment industry.
