When Nvidia showcased humanoid robots at CES in January, crowds marveled at machines that walked, gestured, and appeared to act on their own. But behind the spectacle, as was the case across much of the robotics industry, human operators were quietly pulling the strings. A new investigation from MIT Technology Review, published on February 23, reveals the extent to which the humanoid robot industry depends on concealed human labor — from remote teleoperators wearing VR headsets to armies of low-wage data collectors performing repetitive physical tasks to train the machines.
The report arrives as humanoid robots are entering a pivotal commercial moment. Companies including 1X Technologies, Tesla, and others are racing to deploy robots in homes and factories, with billions in investment flowing into the sector. Yet many of the robots marketed as autonomous still rely heavily on human intervention for even basic tasks.
The Teleoperator in Your Living Room
The Norwegian-American company 1X has been among the most candid about its approach. Its Neo home robot, available for pre-order at $20,000 or $499 per month with deliveries planned for 2026, uses a system called “Expert Mode” in which remote operators wearing Meta Quest 3 VR headsets take control of the robot to complete complex household chores. During a Wall Street Journal demonstration, the robot was unable to complete any task independently — every action, from fetching a water bottle to folding a sweater, required human oversight.
The privacy implications are immediate. Operators can see through the robot’s dual cameras directly into customers’ homes in real time. 1X says it offers controls such as face blurring, room geofencing, and user-approved connection windows, but researchers and privacy advocates remain uneasy. As Eduardo Sandoval, a social robotics researcher at UNSW Sydney, wrote in The Conversation last week, these products are “launched with great fanfare and limited capabilities, concealed privacy risks, and invisible remote workers behind the scenes”.
A Robot Gig Economy Takes Shape
The investigation highlights a broader pattern across the industry: workers, often in low-wage countries, are being deployed at scale to generate the physical demonstration data that robots need to learn. At Tesla, data collection operators have worn motion-capture suits and VR headsets for seven-hour shifts, earning between $25 and $48 per hour, to teleoperate Optimus robots and teach them human movements. Workers told Business Insider that the role caused back and neck injuries, and that prolonged VR headset use triggered severe motion sickness.
The dynamic mirrors well-documented exploitation in AI content moderation and data labeling, where workers in the Philippines, Kenya, and elsewhere review disturbing content or annotate data for a fraction of a living wage. Robotics teleoperation extends this model into physical labor: operators in developing countries could soon be controlling robots in wealthy households, a form of “global wage arbitrage” that one analysis described as potentially creating “a new kind of robot watching” job. In Japan, Filipino operators already remotely monitor and control robots in convenience stores, earning $250 to $315 per month.
The Gap Between Promise and Reality
In January, 1X announced a new “World Model” software update that it says will reduce reliance on teleoperation, allowing Neo to learn from internet-scale video data rather than solely from human demonstrations. A company spokesperson told Business Insider the update “significantly reduces” the need for teleoperators. But experts caution that the gap between controlled demos and reliable home autonomy remains wide, and that teleoperation will be the operational norm for years to come.
Click this link to read our important and interesting articles as soon as they are published!
Ref – https://www.articsledge.com/
