A humanoid robot walks into your facility, and your safety officer asks which standard covers it. The honest answer in 2026 is: several, partially, and one that is still being written.
That gap is not a technicality. It determines your insurance, your liability and whether you can deploy at all.
The Principle Everyone Gets Wrong
Robot safety standards care about the application and the environment, not the visual shape of the machine. A humanoid form does not automatically place a machine outside existing rules, and it does not automatically place it inside a new one either.
What matters is what it does, where it does it, and who is nearby. A legged robot doing inspection in a fenced area is a different regulatory problem from the same robot working beside people.
What Applies Today
| Standard | Covers | Relevance to humanoids |
|---|---|---|
| ISO 10218 | Industrial robots and their integration | The established baseline |
| ISO 3691-4 | Driverless industrial trucks | Applies to wheeled mobile robots |
| Machinery regulations | General machinery safety | Always in scope |
| Occupational safety rules | Worker protection | Applies regardless of the machine |
So you are not operating in a vacuum. You are operating with standards designed before machines that balance dynamically on legs became a realistic proposition.
What ISO 25785-1 Adds
ISO 25785-1 is being developed specifically for dynamically stable industrial mobile robots, covering machines with legs, wheels or other locomotion. Its core concern is the risk posed by loss of stability, which is precisely the gap in existing standards. A wheeled robot that stops is stable. A walking robot that stops mid-stride may not be.
Reporting through early and mid 2026 describes it as a working draft rather than a published standard. Treat it as direction of travel, not as a certificate you can obtain today.
What to Do Before It Lands
- Do a proper risk assessment now. Not a form, a genuine assessment of what happens when the machine falls, misjudges or loses power mid-motion.
- Define the operating envelope. Where the robot may go, at what speed, and who may be there. Then enforce it physically, not just in software.
- Plan for loss of stability specifically. This is the whole reason the new standard exists. What is the fall radius and what is inside it?
- Talk to your insurer early. They will ask which standard applies before you deploy, and an answer prepared in advance is worth having.
- Document everything. When the standard publishes, the organisations that assessed carefully will be close to compliant. The others will start over.
The Commercial Reality
Standards immaturity is a real brake on adoption, and it is one reason serious deployments so far are inside organisations that also build the technology or can carry the risk themselves.
If a vendor tells you their robot is certified, ask certified against what, by whom, and for which application. Vague certification claims in an immature standards environment should be treated as a sales statement rather than a fact.
Conclusion
Existing standards for industrial robots and mobile machinery apply now, judged by application and environment rather than by the machine shape. ISO 25785-1 will close the stability gap and remains a working draft, so build your risk assessment around loss of stability today, define and physically enforce the operating envelope, and involve your insurer before you sign anything. Careful documentation now is most of the compliance work later.
Frequently Asked Questions
Can we deploy a humanoid legally today?
In many jurisdictions yes, with a proper risk assessment and appropriate safeguarding under existing machinery rules. Legal and advisable are different questions.
Does a fenced area solve the problem?
It solves most of it, and it also removes the main reason to buy a humanoid. If the robot works behind a fence, purpose-built automation is usually cheaper and faster.
When will ISO 25785-1 be published?
No firm public date worth quoting. Track it through your national standards body rather than through vendor announcements, which tend to be optimistic.