Such robots are marketed as autonomous but rely on remote operators and low-wage data workers for training and control, investigations reveal.

The emerging humanoid robot industry is far more dependent on hidden human labor than glossy demos suggest, according to a new Massachusetts Institute of Technology (MIT) Technology Review investigation.

The report, published 23 February 2026, describes how robots that appear autonomous onstage or in marketing videos are frequently steered in real time by remote human operators, while others depend on large, low‑wage workforces to collect the training data they need.

This scrutiny comes just as firms such as 1X Technologies and Tesla race to commercialize humanoids for homes and factories, buoyed by billions of dollars in investment and claims that the machines can perform everyday tasks with little human help.

  • 1X has been unusually transparent about this reliance. Its US$20,000 home robot slated for 2026 launch, features an “Expert Mode” in which remote workers wearing Meta Quest 3 headsets guide the robot through household chores. In a demo video, Neo had failed to complete basic tasks without human control, requiring intervention for actions as simple as picking up a water bottle or folding a sweater. The arrangement raises immediate privacy questions, since teleoperators can see directly into customers’ homes through the robot’s cameras. 1X says it uses tools such as facial blurring, geofenced areas, and user‑approved connection windows to limit exposure, but researchers argue these measures do not erase the risks.
  • Social robotics scholar Eduardo Sandoval has warned that such products are launched “with great fanfare and limited capabilities,” while masking both privacy concerns and the invisible workers behind the machines.  The investigation also documents how a global “robot gig economy” is taking shape.
  • At Tesla, operators wearing motion‑capture suits and VR headsets have spent seven‑hour shifts teleoperating its robots, earning between US$25 and US$48 per hour while reporting neck and back pain and severe motion sickness.
  • In some lower‑income countries similar models are emerging, echoing earlier patterns in content moderation and data labeling, where workers in places like the Philippines and Kenya do grueling work for very low pay. In Japan, Filipino staff already remotely supervise convenience‑store robots for only a few hundred dollars per month.

While 1X says a new “World Model” update will let Neo learn more from internet‑scale video and reduce reliance on teleoperation, experts doubt robots will achieve reliable home autonomy anytime soon.

One robotics researcher even noted in response to the MIT Technology Review findings that many eye‑catching demos still depend on extensive behind‑the‑scenes human control and cleanup.