Humanoid Robot Safety Standards Have a Domestic Problem

A humanoid robot can look convincing in a demo and still be unsafe in a home.

That is the gap now opening under the robotics market. Walking, grasping and navigation are no longer the only tests that matter. If humanoid robots are going to enter kitchens, bedrooms, care homes and ordinary apartments, the harder question is whether safety standards can prove that their behavior remains acceptable around real people.

The easy reading is that domestic robotics is waiting for better hardware and better AI. The harder reading is that home robots are becoming a standards problem. Humanoid robot safety standards have to move from machine behavior in controlled spaces toward human-robot behavior inside messy domestic life.

That is why ISO 13482, the international safety standard for personal care robots, matters. It is not a glamorous part of the robot story. It does not have the theater of a humanoid walking onstage, folding laundry or recovering from a shove. But standards often decide when a technology becomes boring enough to sell widely.

For domestic humanoids, boring is the goal.

Why humanoid robot safety standards now matter

The current robot conversation is overloaded with capability. Can the robot walk without falling? Can it pick up soft objects? Can it learn from video? Can a foundation model turn language into action?

Those questions matter. Vastkind has already argued that the future of robotics will be decided by reliability, not robot theater. Reliability is still the first wall between a lab prototype and a useful machine.

Safety is the second wall.

A factory robot can often be separated from humans, fenced into a cell, or limited to a defined workflow. A warehouse robot can be assigned lanes, speed limits and operating zones. A domestic robot is different. Its whole value proposition is that it operates where the environment is not fully controlled.

That means the robot has to behave safely around children, older adults, visitors, pets, loose clothing, dropped objects, wet floors, tight hallways, poor lighting and unpredictable human choices. These are not edge cases. They are the home.

IEEE Spectrum recently framed the issue through the revision of ISO 13482, the personal care robot standard. The article’s key point is not that robot safety has been ignored. It is that the next phase of domestic robots forces standards bodies to confront a harder kind of safety: not only the hazard created by the machine, but the hazard created by the relationship between the machine and the person.

That distinction sounds abstract until you put the robot in a living room.

A home is not a factory

Industrial robot safety can often begin by defining the work cell. The task is bounded. The floor is mapped. The worker population is trained. The objects are known. The workflow is repeatable enough that engineers can test for foreseeable failures.

A home refuses that neatness.

A domestic humanoid might be asked to carry groceries through a hallway while a child runs across its path. It might assist an elderly person whose gait changes over time. It might move near a dog, a walker, a charging cable, a glass table or a sleeping person. It might interpret a spoken request from someone who is tired, confused or physically impaired.

The point is not that domestic robots are impossible. The point is that their safety envelope is not only mechanical. It is behavioral.

A robot does not merely act inside the home. It changes the home. People step around it, depend on it, trust it, ignore it, overestimate it, block it, rush it or try to help it. The human changes what the robot sees and does next. The robot changes what the human does next.

That feedback loop is the domestic problem.

It is also why simple demo metrics are not enough. A robot that completes a task in a staged apartment has not proven that it can remain safe across the range of human states that real homes contain. The safety question is not only, “Did the robot avoid contact?” It is also, “What did the robot cause the human to do?”

The standards gap is relational safety

ISO 13482 gives personal care robots a standards baseline. The public ISO page identifies the standard as covering safety requirements for personal care robots. The full standard is not openly available, which limits public analysis, but its role is clear: it is one of the documents that helps define acceptable safety for robots designed to operate near people outside industrial settings.

The issue raised by IEEE Spectrum is that a revision can acknowledge hazards without fully converting them into enforceable tests.

That matters because domestic humanoid risk is not only a list of machine outputs. Speed, force, stopping distance, stability and collision avoidance are important. But they do not exhaust the problem. A home robot can be physically gentle and still be unsafe if it makes poor autonomous decisions around a vulnerable person.

The harder standard would ask different questions.

How should a robot behave when a person reaches for support and then loses balance? How should it respond when a child treats it as a toy? How should it decide when not to act? How should safety be tested when the same movement is harmless for one adult but dangerous for another?

This is where domestic robot safety becomes political, not in the party sense, but in the standards sense. Someone has to define the reference human.

Whose gait counts as normal? Whose reaction time? Whose home layout? Whose disability? Whose level of confusion, strength or attention? If the standard silently assumes a healthy adult in a tidy room, it may certify a robot for a world that does not exist.

That is the governance layer hiding inside the engineering problem.

Why this matters

Humanoid robots will not reach ordinary homes by winning the demo cycle. They will need evidence that manufacturers, insurers, regulators, retailers and buyers can understand.

That evidence has to answer three practical questions.

First, what has the robot been tested against? A robot tested only in clean labs and scripted apartments is not the same product as a robot tested across domestic clutter, mobility differences and unpredictable household behavior.

Second, who is responsible when the robot behaves within its design limits but the human relationship around it produces harm? A manufacturer can define intended use. A household often produces unintended use by default.

Third, what kind of safety proof can buyers trust? Most people do not read standards documents before buying a device. They rely on certification marks, product norms, reputation, insurance requirements and retailer confidence. Standards translate technical risk into social permission.

That is why the standards process can shape the market before consumers ever see the clause language.

A domestic humanoid company with a better demo may win attention. A company with better safety evidence may win the right to enter homes, care settings and public procurement channels. The winners may not be the robots that look most human. They may be the robots that can prove, in boring detail, how they fail.

This also changes how investors and builders should read the category. Robotics progress is not only model progress. Large behavior models may change robotics’ real bottleneck, but behavior still has to be bounded, tested and certified. Autonomy without standards becomes a liability machine.

The market does not need humanoids that appear safe onstage. It needs humanoids that can be inspected.

What remains unproven

The evidence boundary is important here.

It is proven that ISO 13482 exists as the relevant personal care robot safety standard. It is also clear from IEEE Spectrum’s reporting that experts are focused on how the standard handles domestic humanoid deployment, human-robot interaction and the limits of advisory guidance.

It is not yet proven that the revised standard will become a hard market-access threshold. It is not yet proven that insurers or regulators will treat it as the decisive line. It is not yet proven that domestic humanoid manufacturers will publish enough safety evidence for outside observers to compare them seriously.

It is also too early to claim that standards are blocking the market today. Hardware cost, battery life, dexterity, reliability, repair, support and business models still matter. Home robots have many bottlenecks.

The claim is narrower and stronger: as domestic humanoids move closer to real homes, safety standards become part of the product. Not paperwork after the fact. Part of the product itself.

That is the difference between a device that can perform a task and a device that can be trusted near a person.

The market test is trust, not theater

The next phase of domestic robotics will be judged badly if it is judged only by videos.

A video can show a humanoid making breakfast. It cannot show the distribution of near misses across thousands of kitchens. It cannot show whether an older adult changed behavior because the robot was present. It cannot show whether the robot recognizes when a task should be refused. It cannot show whether a safety case survives contact with a crowded home.

That is why standards are not a side story. They are one of the places where robot capability becomes social reality.

The same lesson appears in other parts of the robotics cluster. A humanoid robot marathon turns performance into a reliability test because endurance exposes weaknesses that demos hide. Domestic safety does something similar. It forces robotics companies to prove not only that machines can act, but that they can act within human environments without shifting hidden risk onto the people around them.

This does not mean domestic humanoids are doomed. It means the credible version of the market will be slower, more constrained and more institutional than the demo cycle suggests.

The companies that understand this will treat safety evidence as core infrastructure. They will design for testing, publish clearer limits, work with standards bodies and avoid pretending that a home is a showroom.

The companies that do not will keep producing impressive clips and fragile trust.

Domestic humanoids do not need only better hands, eyes and batteries. They need a credible way to prove that their behavior remains safe around real people. Until that exists, the home is not an addressable market. It is an unresolved test.

For readers tracking the broader robotics shift, the next useful step is the Robotics hub, where Vastkind follows the move from robot spectacle toward deployment, reliability and social consequence.

Disclosure: This article is editorial analysis of robotics safety standards and market readiness. It is not legal, compliance or product-safety advice.