DoorDash Tasks App Pays Couriers to Film Chores for AI Robot Training in 2026

DoorDash is paying its delivery couriers to strap on body cameras and film themselves washing dishes, folding clothes, and making beds. The footage trains AI and robotics models. The app launched on March 19, 2026. It's called Tasks, and it's available to DoorDash's 8 million US couriers — except in California, New York City, Seattle, and Colorado.

That exclusion list tells you something the launch announcement doesn't.

What DoorDash Actually Built

Tasks operates on two levels. Inside the existing Dasher delivery app, couriers can pick up small jobs between orders: photographing a restaurant's dishes for its menu, snapping a hotel entrance so future drivers find the drop-off, or scanning supermarket shelves.

The standalone Tasks app is different. It has nothing to do with delivery. Couriers film themselves doing household chores — washing at least five dishes while wearing a body camera, holding each clean dish in frame for a few seconds. They fold clothes. They make beds. They prune and repot plants. They record unscripted conversations in Spanish with their friends and family.

"This data helps AI and robotic systems understand the physical world," DoorDash wrote in its announcement. Pay is shown upfront, weighted by effort and complexity. Harder tasks pay more.

What DoorDash didn't publish: average pay rates, minimum pay floors, data retention policies, or what rights workers have over footage recorded inside their own homes.

The $3 Billion Data Race

DoorDash isn't inventing this market. It's arriving in one that already exists and growing fast. The global data annotation industry is worth roughly $2-3 billion in 2026 and projected to reach $7-14 billion by the early 2030s, depending on whose estimate you trust.

What's new is who's doing the work and where.

Scale AI built a billion-dollar company on remote data labelling — workers at screens, tagging images and text. But AI systems that need to manipulate physical objects can't learn from screenshots. They need real humans doing real tasks in real kitchens. That's embodied data, and it's the scarcest, most valuable training data in AI right now.

DoorDash's insight was that it already owns the logistics: 8 million workers dispersed across almost every postcode in the US, a dispatch system that can assign tasks by location, verification infrastructure, and payment processing. It didn't need to build a data collection workforce. It had one delivering burritos.

Uber saw this too. In October 2025, it launched a similar pilot through its AI Solutions Group, paying drivers to upload photos and recordings. Instawork, a staffing app, went further: the Los Angeles Times reported in March that the company was recruiting workers in LA to strap on headbands with phone mounts and film themselves cleaning their homes.

Sunday Robotics ships a "Skill Capture Glove" to thousands of people across the US. They wear it while doing household tasks, and their hand movements become training data for Memo, the company's home robot, which starts beta deliveries late this year.

The Automation Paradox

Here's the part most coverage doesn't say plainly: the workers filming themselves washing dishes are producing the training data for robots that wash dishes.

DoorDash's co-founder Andy Fang called it "building the frontier of physical intelligence." The humanoid robot market is projected to hit $15 billion by 2030, up from under $3 billion today. Companies like Figure, Tesla, and Sunday Robotics are all racing to build machines that can navigate homes and complete domestic tasks. Every one of them needs exactly the data DoorDash is now collecting at scale.

DoorDash itself already partners with Waymo on a task that makes the loop uncomfortably explicit. When a passenger leaves a Waymo self-driving car door open — a safety issue that prevents the autonomous vehicle from driving away — nearby Dashers get a notification. They drive over and close the door for about $11. Gig workers, manually intervening to keep autonomous vehicles running. Waymo has said future models will have automated door closures.

A March 2026 Longreads investigation put it starkly: "Producing training data means working toward your own obsolescence." New York Magazine found that laid-off professionals — lawyers, scientists, writers — are increasingly hired back as contractors to train the AI systems that replaced them.

DoorDash hasn't framed it this way. The company calls Tasks "a new way to earn on their own terms."

The Privacy Question Nobody's Asking

The Tasks app asks workers to film inside their own homes. To record conversations with their families. To wear body cameras while doing domestic work.

DoorDash has not published details on consent mechanisms, data retention periods, or what happens to footage after it's submitted. According to The Next Web's analysis, the company hasn't disclosed whether workers can request deletion of their footage, who gains access to the audio recordings, or how long the data persists.

Bloomberg reports that the footage trains both DoorDash's in-house models and those of its partners in retail, insurance, hospitality, and technology. That means a courier's kitchen footage could end up training models for companies the courier has never heard of.

The jurisdictions excluded from Tasks — California, New York City, Seattle, Colorado — all have stronger gig worker protections or data privacy regulations than the rest of the country. California's consumer privacy laws give residents the right to know what data companies collect and to request its deletion. Colorado's Privacy Act requires consent before processing sensitive data. DoorDash's decision to avoid these markets suggests the company knows its current data practices wouldn't survive tighter scrutiny.

Why This Story Is Only American

This story has been covered exclusively by US outlets: Bloomberg, TechCrunch, NBC News, Forbes, CNET, The Next Web. No major European, Asian, Middle Eastern, or Latin American outlet has picked it up.

That's partly because DoorDash operates primarily in the US. But the implications aren't American. The data annotation industry employs hundreds of thousands of workers globally — many in Kenya, India, and the Philippines, where companies like Scale AI and Appen run large labelling operations. The model DoorDash just formalised — turning a logistics workforce into a data collection workforce — is exportable to any gig platform with enough drivers.

Uber's AI Solutions Group piloted in India before launching in the US. Grab, GoTo, and Meituan operate delivery networks across Southeast Asia and China with millions of couriers. If embodied data collection becomes standard practice for gig platforms, the workers most affected won't be in San Francisco. They'll be in Manila, Nairobi, and Mumbai.

What Happens Next

DoorDash says Dashers have completed more than two million tasks since 2024, before the formal launch. The company plans to expand into more task types and countries.

The data annotation market is projected to grow at 27-32% annually through the decade. The humanoid robotics market is growing even faster. Every one of these systems needs the kind of physical-world training data that DoorDash is now mass-producing through its courier network.

Two million tasks is a pilot. Eight million potential data workers is a platform. The question isn't whether this model scales — DoorDash already proved the logistics work. The question is what the workers filming their kitchens today will do when the robots they trained don't need them anymore.

DoorDash's answer, for now, is that there will always be more tasks. History suggests that's true right up until there aren't.