On October 13, 2026, South Korea's Ministry of Trade, Industry and Energy announced that it had approved a regulatory exemption for a demonstration in which robot developer ROBROS will deploy 20 humanoid robots at industrial sites. The trial covers four industries: semiconductors, automobiles, shipbuilding and logistics. The robots are planned to operate in the same space as workers or in adjacent areas. Safety standards built around fixed robots have made it difficult to introduce machines that walk while they work. The demonstration, however, requires a risk assessment by labor authorities and measures to reduce hazards. This step moves the question from whether fences can be removed to how safety can be maintained in specific processes.
Four industries test work capability, not just walking
The robots ROBROS will deploy are AI-equipped bipedal humanoids. Unlike demonstrations that simply show a robot walking, these machines are meant to perceive their surroundings, grasp objects, and carry out transport and loading. The ministry's detailed materials say the demonstration will verify the ability to perform tasks and adaptability to the work site.
According to the ministry's materials dated October 13, 2026, the demonstration is a conditional verification that combines the deployment of 20 units across four industries with risk assessment by labor authorities and hazard-reduction measures.
| Industry / Company | Planned task | Target of the work |
|---|---|---|
| Semiconductors / Nepes | Removing dirt from parts | Circular CRS parts used in the back-end semiconductor packaging process |
| Automobiles / Sungwoo Hitech | Placing body parts in position | Automobile body parts |
| Shipbuilding / HD Hyundai Samho | Loading parts | Parts that secure scaffolding for high-place work to ship hulls |
| Logistics / Delivery Lab | Loading packages | Packages used in cold-chain logistics |
The source is pages 2–3 and 9 of the ministry's announcement. The tasks have been sorted by industry, with the shared conditions of the demonstration noted alongside. The figure of 20 units is the overall scale; it is not a breakdown by site or a success rate for any task.
The cleaning task in the semiconductor setting is not part of the process of building circuits on wafers. It is the removal of dirt from parts used in the back-end packaging process. Likewise, the robots will not take over the entire assembly of a ship; the target is the loading of parts used to secure scaffolding. Looking closely at each process clarifies what capabilities are being tested and which results cannot yet be claimed.
Even among humanoids, removing dirt and placing parts change what counts as success. Beyond being able to move, the robots will likely be judged on whether they can handle objects properly and meet each process's requirements. The ministry says it will expand the demonstration from simple transport to complex processes involving precision cleaning and low-temperature logistics, which is a distinctive feature of the plan.
Exceptions to safety fences already existed
Article 223 of South Korea's Occupational Safety and Health Standards Rules generally requires fences at least 1.8 meters high. However, the text published by the Korea Ministry of Government Legislation's National Law Information Center also states that the height can be adjusted if there is no danger, taking into account factors such as the robot's range of motion, and that for some sections where fences cannot be placed, such as conveyors, safety mats or photoelectric protective devices must be provided.
There is also a proviso: if the Minister of Employment and Labor recognizes that a robot's safety standards conform to Korean Industrial Standards or internationally accepted safety standards, the measures in the main text of the article can be omitted. In other words, the existing system already allowed exceptions depending on confirmed safety. Interpreting this as "in Korea, every robot had to work inside a fence" would overstate the change.
There is also a history of measures for mobile robots. On July 4, 2023, the Ministry of Employment and Labor and the Korea Occupational Safety and Health Agency published a guide for the safe use of mobile industrial robots. Along with safety measures such as sensors, it compiled ways for workers to move and work safely, and included related standards, self-inspection checklists and usage examples. Even then, it anticipated the use of robots that collaborate with people while moving.
However, the existence of this guide alone does not mean the bipedal robots in this trial have been recognized as meeting the necessary safety requirements. The ministry explains that conventional industrial robot safety standards are not suited to humanoids that move freely while working, making them hard to use on site. It has therefore granted the exemption for ROBROS's specific demonstration on the condition that labor authorities conduct a risk assessment and that hazard-reduction measures are taken.
This is a stage in which a new form of robot is tested at work sites while building on existing safety measures. It is neither a blanket permission to remove fences at factories nationwide nor an announcement that general safety standards for humanoids have been completed.
Next question: the conditions for maintaining safety on site
The ministry says it will use the demonstration to secure data for improving safety standards and systems related to robot use. Because the targets are tasks in the same space as workers or in adjacent areas, the results will be hard to evaluate on a robot's dexterity alone.
For example, the ability to handle objects in a process and the ability to limit danger when a person approaches are questions best verified separately. Even if a robot can place a part in the designated position, the conditions under which it can be built into a process are limited if it cannot resume work after stopping for people nearby. Conversely, if it stops every time someone approaches, the hazard-limiting mechanism and the progress of the work must be evaluated together. This is a viewpoint for reading future results; the authorities have not announced any specific stopping method or evaluation metric.
What the current materials leave unpublished includes travel speed, the weight of objects handled and task success rates. It is also unclear how much human intervention is needed, or when the demonstration will start and how long it will last. From a scale of 20 units, labor-saving effects or worker displacement cannot be discussed in numerical terms. Between being given a chance to test and being able to keep using the robots in a factory, verification still remains.
For those considering adoption, what matters is for which tasks, and with what safety measures and human involvement, operation proved possible. If those conditions are shown along with concrete data, other factories will find it easier to judge, in light of their own processes, whether they can accept humanoid robots.