A humanoid robot is built to work in spaces where people are present too. That is its greatest promise and, at the same time, its greatest challenge. An industrial robot arm usually stands behind a fence; a humanoid is meant to walk through corridors, open doors and hand objects to people. This article looks at the risks involved, the standards that exist and what safety looks like in practice.
Why humanoids are different
Most robot safety rests on a simple principle: in case of danger, the machine stops. A robot arm or a wheeled warehouse robot simply stays put after an emergency stop. A humanoid on two legs does not work that way. It only stays balanced by continuously correcting itself. If the power is cut, the robot falls over.
i-SCOOP sums it up: a conventional mobile robot stays where it is when switched off; a humanoid does not. That has consequences for maintenance and emergency procedures. The classic "power off and lock out" can itself create a hazardous situation with a humanoid.
Other differences:
- Mass and height. An adult-sized humanoid easily weighs 30 to 70 kilograms or more. A fall or collision can cause injury.
- Many joints. With dozens of joints, there are more places where fingers or clothing can get caught.
- AI-driven behaviour. Robots that use AI models behave less predictably than a robot running a fixed program.
- Changing environments. In a home or shop, conditions are always different, with children, pets and loose objects.
The main risks
- Collision and impact: the robot strikes a person with an arm, hand or its body.
- Pinching and crushing: fingers or clothing get trapped between moving parts.
- Falling: the robot loses its balance and falls on or next to a person.
- Dropped objects: the robot drops something it is carrying.
- Battery and electrical safety: overheating, fire or charging problems.
- Cybersecurity: someone takes remote control or gains access to cameras and microphones.
- Privacy: the robot records images and sound of people who did not choose this.
Which standards exist?
ISO 10218: industrial robots (2025)
The revised ISO 10218 was published in 2025, its first major overhaul since 2011. Part 1 addresses manufacturers, Part 2 integrators. The requirements for collaborative robots, formerly ISO/TS 15066, have been incorporated, according to The Robot Report. Well-known protective principles from these standards include:
- Speed and separation monitoring: the robot slows down or stops when a person comes too close.
- Power and force limiting: the robot may not exceed a set force on contact.
- Safety-rated monitored stop: the robot stops safely when a person enters the workspace.
However, ISO 10218 assumes that a robot does not tip over. Stability is therefore not covered.
ISO 25785-1: robots with active balance (in development)
ISO 25785-1 is being developed specifically for industrial mobile robots that must actively maintain their balance, such as humanoids, four-legged robots and self-balancing wheeled robots. The standard has not been published yet. According to i-SCOOP, the draft covers fall zones, fall detection, controlled descent during an emergency stop or power loss, and limits on contact forces. Publication is expected in late 2026 or in 2027. A second part on application integration will follow later.
ISO 13482: personal care and service robots
For robots that help people outside the factory, there is ISO 13482. That standard was not written with bipedal humanoids in mind. A French analysis noted in early 2026 that no consumer humanoid had yet been certified to this standard for the EU market (Korben).
Legislation
Alongside standards, EU legislation applies. The Machinery Regulation, applicable from 20 January 2027, requires assessment by a notified body for safety functions that use self-evolving machine learning. From August 2028, the AI Act sets requirements for AI built into machinery as a safety component. Our article on CE marking, the AI Act and the Machinery Regulation in Insights explains more.
According to investor Sam Baker, quoted by Euronews, safety standards are currently the biggest bottleneck for deploying robots alongside workers.
What are manufacturers doing themselves?
Manufacturers take design measures to reduce risk. A few examples, as described by the manufacturers themselves:
- 1X describes NEO as having a soft body made of a polymer lattice, joints without pinch points and low-inertia drives for gentle movement.
- Figure AI uses foam at pinch points on Figure 03 and a battery with multiple safety layers and UN38.3 certification (a transport test for lithium batteries).
- Some manufacturers have the robot kneel first when stopping, to reduce the fall height.
These are the manufacturers' own claims. Independent certification against a dedicated humanoid standard does not yet exist.
What does safety look like in practice today?
Today: in industry, humanoids mostly work in defined or fenced zones, at limited speed and with clear rules on who may approach and when. Home robots are designed to be light and soft, but are used in small numbers and partly under remote supervision.
In development: standards for dynamically balancing robots, reliable fall detection and controlled descent, and methods to test and validate AI behaviour.
Later: humanoids that work alongside people without fencing, with demonstrably controlled risks and independent certification.
Checklist for anyone deploying a humanoid
- Ask the manufacturer which standards have been applied and which safety functions are certified.
- Have it demonstrated what the robot does during an emergency stop and a power failure.
- Carry out a risk assessment for the specific workplace or home, including fall zones.
- Limit speed, force and working area at first, and define who may operate the robot.
- Arrange cybersecurity: updates, access control and what happens when the connection is lost.
- Document which camera and audio recordings are made, and inform the people around the robot.
- Train employees or users, and review incidents and near misses.
In closing
With humanoid robots, safety is not an afterthought but the precondition for wider use. The technology is ahead of the standards, which for now calls for caution and clear agreements. Our Technology page explains more about how humanoids work. Want to stay informed? Sign up for Early Access.
Sources
- https://www.i-scoop.eu/iso-25785-1-explained-and-what-it-means-for-humanoid-robot-safety/
- https://www.iso.org/standard/91469.html
- https://www.therobotreport.com/iso-10218-industrial-robot-safety-standard-receives-major-overhaul/
- https://korben.info/korben-aura-robot-humanoide.html
- https://euronews.com/2026/03/05/is-europe-losing-the-robotics-race-to-china-and-does-it-matter
- https://www.1x.tech/neo
- https://www.figure.ai/news/introducing-figure-03
- https://osha.europa.eu/en/legislation/directive/regulation-20231230eu-machinery
- https://www.bakermckenzie.com/en/insight/publications/resources/product-risk-radar-articles/machinery-regulation
- https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai
This article is based on public sources and reflects the situation on the date of publication. It is not advice on any specific purchase.