Home robot companies ask buyers to pay £16,000 whilst remote operators watch through their cameras
Tesla and 1X admit their humanoid robots need substantial human control, turning customers into unpaid AI trainers
The £16,000 robot assistant arrives in 2026. It will tidy your rooms, load your dishwasher, fetch your drinks. There's just one detail: a stranger thousands of miles away will control it by watching your home through its camera eyes.
When a Wall Street Journal reporter tested 1X Technologies' Neo humanoid robot, the machine didn't perform a single task autonomously. Fetching water took over a minute. Loading three dishes took five. Every movement was guided by someone called "Turing" wearing a VR headset in another room - a human operator whose real name went unreported.
This isn't a prototype glitch. It's the business model.
1X chief executive Bernt Børnich states the terms bluntly: buyers must accept a "social contract." For Neo to work, customers must allow human operators to see inside their houses through the robot's cameras. An app lets you schedule when remote workers take control and set no-go zones. The company can blur faces. But the transaction remains unchanged: strangers pilot a machine through your home to complete tasks the robot cannot do alone.
And 1X isn't unique. The company is simply more honest about what's become standard practice across the entire humanoid robotics industry.
The wizard behind the curtain
At Tesla's "We, Robot" event in October 2024, Optimus robots mingled with crowds, served drinks, played games. The performance looked impressive until observers realised the robots were substantially teleoperated. Wall Street analyst Adam Jonas called the demonstration heavy on "tele-ops" - prioritising spectacle over independence.
Figure AI, backed by Microsoft, OpenAI and Jeff Bezos, uses identical methods. The pitch sounds reasonable: humans control robots to generate training data that eventually teaches AI models to work alone. But timelines keep slipping. Tesla announced Optimus in 2021, promising production by 2023. Mid-2025 arrived with roughly 1,000 prototypes performing simple factory tasks under controlled conditions. Household demonstrations like folding laundry still need remote pilots.
The engineering problems are real. Navigating unpredictable homes, manipulating diverse objects, recovering from errors autonomously - these remain extraordinarily difficult. What's changed isn't the technology. It's the business model companies now deploy whilst waiting for breakthroughs that may never come.
The offshore labour model
Tokyo convenience stores now employ robots that restock shelves. More than 300 locations use them. The robots are controlled from Manila.
Sixty young workers at Astro Robotics in the Philippines monitor roughly 50 robots each, earning between $250 and $315 monthly - around £200 to £250. They intervene only a few percent of the time when robots encounter difficulties. But that human oversight is what makes the system function.
This is the quiet truth about "autonomous" commercial robots: they run on workers in developing countries earning wages no wealthy nation would tolerate. Physical work once immune to offshoring - restocking shelves, cleaning rooms - can now be performed by distant workers teleporting into robotic bodies.
The economics are brutal and obvious. Developing truly autonomous AI costs enormous sums with no guaranteed timeline. Hiring operators in countries with lower labour costs provides cheaper, reliable solutions today. For companies, teleoperation bridges the gap whilst generating training data. For workers, it's employment at wages that make the model viable precisely because they're so low.
1X describes this as temporary. The company's Redwood AI model, announced in June 2025, combines locomotion and manipulation into whole-body control. But the 2026 launch roadmap still includes substantial human oversight, with autonomous work quality varying significantly and improving only as more customer data accumulates - data provided free by people who paid £16,000 for the privilege.
The privacy calculation
A Ring doorbell watches one angle. Alexa listens in one room. Neo walks through your entire house with cameras and microphones, potentially operated by workers thousands of miles away.
The privacy controls sound reassuring - blurred faces, no-go zones, operator approval. They don't change the transaction. To use Neo for tasks it hasn't learned, you allow remote workers to view your home through its sensors. The company says you're "in control." Yet the product only works by surrendering that control during scheduled sessions.
Security researchers documented what happens when these systems fail. A September 2025 study examined the Unitree G1 humanoid robot and found easily hackable Bluetooth encryption, hard-coded keys, and a disturbing discovery: the robot was secretly sending information to China. The researchers concluded it could be used for covert surveillance or to launch cyberattacks.
These aren't hypothetical risks. As humanoid robots proliferate, they install surveillance infrastructure that didn't exist before. Whether exploited by operating companies, governments demanding legal access, or criminals breaching security, the capability is being placed directly into homes. Studies on teleoperated robot privacy found that whilst visual filters reduce some concerns, the physical presence of a remotely operated machine in intimate spaces creates entirely new vulnerabilities. People worry most about where their data goes. But they should also worry about who's watching right now.
Lessons from the robot graveyard
Jibo raised $73 million. The social robot recognised faces, told stories, performed charming dances. Time magazine called it one of 2017's top inventions. By 2018, Jibo was dead. Novelty wore off. Customers couldn't justify $900 for what a £50 Amazon Echo did better.
Kuri followed the same path. Parent company Bosch shuttered the security robot in 2018 after investors declined to fund long-term development. Anki raised $180 million for Cozmo and Vector, sophisticated toys praised for emotional intelligence. The company closed in 2019.
These failures share a pattern: positioning robots as companions that would provide sustained engagement, then failing to deliver ongoing value justifying premium prices. The personality wore thin. Without utility, expensive novelty couldn't survive.
Roomba succeeded by avoiding this trap entirely. The vacuum doesn't try friendship. No expressive face. No charming voice. It cleans floors, returns to its dock, finishes. Since 2002, iRobot has sold millions by focusing on one task done reliably rather than emotional connection.
The question facing Neo is whether offshore labour changes this equation. If companies keep operator costs low enough whilst gradually improving autonomy, they might sustain businesses without achieving the AI revolution they advertise. That creates permanent dependency on remote workers rather than technological breakthrough.
Consider the adoption curve. Roomba started around £200, proved its value, then introduced premium models. Neo asks £16,000 upfront and years training an AI that may never work independently. Early adopters aren't buying finished products. They're funding development.
What buyers should understand
Pre-ordering Neo means understanding precisely what you're purchasing. Not a fully autonomous assistant. A hybrid system combining limited autonomy with scheduled remote operation. Workers in distant locations controlling a camera-equipped machine through your home. Contributing unpaid labour to train AI models that remain company property.
The regulatory landscape hasn't caught up. Consumer protection laws weren't written for teleoperated home robots. Liability when remote operators cause damage remains unclear. Labour standards for overseas pilots are poorly defined. Europe's AI Act took effect in January 2025, mandating transparency and consent for data-collecting devices, but enforcement barely exists.
Companies face genuine technical challenges. Creating affordable, capable, safe humanoid robots that navigate real homes and perform diverse tasks autonomously may require decades more work. Teleoperation provides a pragmatic path to deployment whilst collecting training data. From a business perspective, it makes sense.
But buyers deserve honesty about what they're getting. When companies market robots as autonomous whilst substantially relying on human operators, they misrepresent their products. When they ask premium prices whilst requiring customers to surrender intimate privacy and perform unpaid training work, they propose transactions demanding scrutiny rather than enthusiasm.
The home robot future may arrive. The question is whether it arrives through AI breakthroughs delivering autonomous systems worth their cost, or through business models exploiting cheap offshore labour whilst conditioning consumers to accept surveillance for convenience. Right now, the industry bets on the latter whilst advertising the former.
1X ships Neo in 2026. Tesla aims to produce thousands of Optimus units in 2025 for factory use before broader sales in 2026. Figure AI projects household trials late 2025. We'll discover whether these companies deliver on promises, or join Jibo, Kuri, and Anki in the graveyard of home robots that couldn't sustain the dream.