Last updated: October 22, 2026

Humanoid robot privacy is the question of what a robot's cameras, microphones and depth sensors capture inside a home or workplace, where that data goes, and who can see it. A humanoid is a networked sensor platform on legs: it moves from room to room, it may be watched by a remote human operator, and its footage is often used to train the next version of its AI. This guide explains the real privacy risks of home robots, what the rules say, and what to check before you let one in.

TL;DR
  • Humanoids collect far more than smart speakers: video, audio, depth maps and a floor plan of your home, often sent to the vendor's cloud.
  • Three real paths to exposure: vendor telemetry, remote teleoperators, and training data handled by people you never meet.
  • Ask for local processing, an operator-access log, opt-out of training, MFA and a clear deletion policy before you buy or deploy.

What data does a humanoid robot collect?

A humanoid needs to perceive its surroundings to walk, grasp and avoid people, so the sensors are always doing something. Depending on the model, that typically includes:

  • Cameras (often several, including stereo or fisheye) for navigation, object recognition and teleoperation.
  • Microphone arrays for voice commands and, on some models, conversation with a built-in assistant.
  • Depth and LiDAR sensors that build a 3D map of rooms, furniture and doorways.
  • Location and motion data: where the robot went, when, and what it touched.
  • Interaction history and "memory" of conversations and tasks.

The result is closer to a mobile CCTV system with a microphone than to a smart speaker. As we explain in our humanoid robot attack surface map, the same sensors that make the robot useful are also the richest target on the device.

Do humanoid robots spy on you?

Not by design in most cases, but there is published evidence that some robots send more data, more often, than owners are told. In September 2025, Alias Robotics published a security assessment of the Unitree G1 humanoid (arXiv 2509.14139). As reported by Help Net Security, the researchers observed the robot sending telemetry, including data from cameras, microphones and internal services, to servers located in China every five minutes, starting within seconds of power-on, with no visible indicator or consent prompt. The article did not include a response from Unitree. We cover the separate G1 remote code execution flaws in Unitree G1 vulnerabilities explained.

"Spying" is usually the wrong frame. The more common risk is ordinary data collection that is poorly disclosed, poorly secured, or reused for purposes the owner did not expect.

The three ways home robot data gets exposed

1. Vendor telemetry and cloud storage

Most consumer robots rely on a cloud account for updates, maps and app control. Anything stored there can be breached, subpoenaed, or accessed by staff. The closest precedent is the home camera market: in 2023 the US Federal Trade Commission said that Ring had given employees and contractors access to customers' private videos, that one employee viewed thousands of recordings of female users, including in bathrooms and bedrooms, and that hackers took over about 55,000 customer accounts. Ring agreed to pay $5.8 million, to delete data, models and algorithms derived from videos it unlawfully reviewed, and to require multi-factor authentication. A humanoid with a camera in every room carries the same risk, plus the ability to move.

2. Remote teleoperators

Many humanoids still need humans in the loop. For the 1X NEO home robot, Engadget reported that a remote operator can see inside the home through the robot's camera for tasks NEO cannot yet do alone, and that buyers must accept this. According to the same report, owners schedule and approve these sessions, people can be blurred, and owners can set no-go zones. 1X's own NEO page describes owners scheduling "a 1X Expert" to guide the robot through unknown tasks. Those controls are a reasonable start; the open questions are logging, retention, and who audits the operators.

3. Training data and human labelers

Footage from homes is valuable training data, and 1X's CEO told Engadget plainly: "If we don't have your data, we can't make the product better." The risk is what happens next. In 2022, MIT Technology Review revealed that images captured by development versions of iRobot's Roomba J7, including a woman on a toilet, were sent to data-labeling contractors through Scale AI and leaked to private social media groups. iRobot said the devices were special development units used by paid collectors who had signed consent agreements, and that the images were shared in breach of a non-disclosure agreement. The lesson for humanoids: once video leaves the home for annotation, the number of people who can see it grows quickly.

What the law says about robot data privacy

Rules differ by region, and none were written specifically for humanoids. The main ones that apply today (this is not legal advice):

Rule Where What it means for robots
GDPR EU/EEA Video and audio of identifiable people is personal data. Requires a lawful basis, data minimisation and data protection by design (Article 25).
EU AI Act EU Bans AI systems that infer emotions of people in the workplace, except for medical or safety reasons (CDT summary). Relevant to humanoids with "empathetic" assistants in offices or factories.
FTC Act, Section 5 US The FTC treats misleading privacy claims and lax security as unfair or deceptive, as in the Ring order.
FCC Covered List US Targets foreign-made robots on national security grounds, partly over data flows. See FCC robot ban explained.

Security standards matter here too: privacy promises are only as good as the device security behind them. Our guide to humanoid robot security standards covers IEC 62443, ISO 10218 and the EU rules.

Is it safe to have a humanoid robot at home? A privacy checklist

Analysis: a home humanoid can be reasonably private, but only if the vendor gives you real controls and you use them. Before buying or deploying, check:

  1. Local processing. Which perception runs on the robot, and what is uploaded? Ask for a data-flow list, not a marketing summary.
  2. Visible indicators. Is there a hardware light or shutter when cameras and microphones are recording or streaming? Researchers at SOUPS 2024 tested 86 ways for domestic robots to signal sensor status (Windl et al.); insist on at least one you can see.
  3. Operator access. Can a remote human ever see or control the robot? Is every session approved by you, logged and viewable afterwards?
  4. Training opt-out. Can you stop your footage being used to train models, and is it blurred or reviewed in-house before anyone sees it?
  5. Retention and deletion. How long is video kept, and can you delete it, including derived maps and models?
  6. Account security. Multi-factor authentication on the owner account and on vendor staff access.
  7. Network control. Put the robot on its own network segment and watch where it connects. Unexpected destinations are a red flag.
  8. Updates and disclosure. Does the vendor publish security advisories and fix flaws? Check our Humanoid Robot Vulnerability Tracker.

For businesses, our free Humanoid Robot Security Checklist turns these into procurement questions.

Frequently asked questions

Can a humanoid robot record me without my knowledge?

Yes, it is technically possible. A robot's cameras and microphones run for navigation and voice control, and research on the Unitree G1 found telemetry being sent without a visible indicator. Check for recording lights, logs and a clear data policy.

Can a remote operator see inside my home?

On some models, yes. Teleoperated humanoids such as the 1X NEO let a remote human see through the robot's camera for certain tasks. Reported controls include owner approval, scheduling, blurring of people and no-go zones.

Is my robot's video used to train AI?

Often. Vendors use real-world footage to improve their models. Look for an opt-out, on-device blurring and in-house review, because outsourced labeling has leaked home images before.

What happens to my robot data if it is hacked?

Anything the robot can see or hear, an attacker may be able to access. Read can humanoid robots be hacked? for documented cases.

Which privacy laws apply to humanoid robots?

In the EU, GDPR and the AI Act; in the US, mainly FTC consumer protection law plus state privacy laws. None are humanoid-specific yet.

Sources

Get the weekly Humanoid Threats Brief → Subscribe