Robotics Field Guide
Topic hub

Safety and standards

What the main robot safety standards cover and how they fit together. Risk assessment, industrial robots and cells, collaborative applications, service robots, safety-related control systems and robot security.

Updated 14 Nov 2026 · robot safety, risk assessment, ISO 12100, ISO 10218, ISO/TS 15066, collaborative robot, ISO 13482, ISO 13849, IEC 62061, performance level, functional safety, IEC 62443, robot security

Overview

A robot that can move a tool at speed near people needs a safety case before it goes into service. That case rests on a risk assessment for the whole application, safety functions built to a known level of reliability, and, more and more, protection against tampering over the network. International standards set out how to do each of these. This hub explains what each of the main robot standards covers, who it applies to and how they connect. It is for engineers who integrate robots, write robot software or have to talk to a safety engineer.

Key ideas

Risk assessment, the robot and the cell

Every robot safety standard sits on top of a risk assessment. ISO 12100:2010 gives the general method for machinery: identify hazards across the machine’s life cycle, estimate and evaluate the risk, and reduce it. Its strategy for risk reduction works through inherently safe design first, then safeguarding and complementary measures, then information for use. ISO lists it as under review at the time of writing.

define theapplicationidentifyhazardsestimateriskevaluate:acceptable?documentand validateyesnoReduce risk, in this order1. inherently safe design2. safeguarding, complementary measures3. information for usere-assess:new hazards?
A simplified risk assessment loop for a robot application, in the spirit of ISO 12100. Each risk reduction step can introduce new hazards, so the loop returns to hazard identification. This sketch is ours and is not a figure from the standard.

The ISO 10218 series applies this to industrial robots. ISO 10218-1:2025 covers the robot itself, treated as partly completed machinery: its inherently safe design, protective measures and information for use. ISO 10218-2:2025 covers integrating robots into applications and robot cells, from design through commissioning, operation and maintenance. The 2025 editions replace those of 2011. According to ISO’s catalogue page, the changes include clarified functional safety requirements, two classes of robot for functional safety, requirements on security where it affects safety, and collaborative-application content that earlier sat in ISO/TS 15066. The practical point is that a robot built to Part 1 does not make a safe cell. The tool, the workpiece, the layout and the people around it all enter the assessment for Part 2.

Collaborative applications

ISO/TS 15066:2016 is a technical specification for collaborative industrial robot systems and their work environment, written to supplement ISO 10218. It is commonly summarised through four modes of collaboration: safety-rated monitored stop, hand guiding, speed and separation monitoring, and power and force limiting. Its ISO page still lists it as published, marked for revision. A cobot is a robot designed for such applications. It is not safe by itself. A cobot holding a knife or a hot part is a hazard, whatever its force limits, so collaboration is a property of the application and its risk assessment.

Personal and service robots

ISO 13482:2014 sets safety requirements for personal care robots that work close to people outside industry, and covers three types of such robot. A revised edition, retitled for service robots in personal and professional use, is in progress as ISO 13482. At the time of writing we could not confirm whether it has been published, so check its ISO page.

Safety functions and security

A safety function, such as a protective stop when a light curtain is broken or a speed limit when a person is near, must work with a known reliability. Two standards cover how to design and integrate the control systems that carry such functions. ISO 13849-1:2023 gives a method for safety-related parts of control systems, including software, and rates them by performance level. It does not tell you which safety functions an application needs or which level each requires. That comes from the risk assessment and from product standards such as ISO 10218. IEC 62061, edition 2.0 of 2021 with Amendment 1 of 2024, covers the same ground for machinery as a sector standard under IEC 61508.

Robot security is now part of safety. A controller on a factory network, a teach pendant, or a ROS 2 graph left open to any node can be used to change a safety-relevant setting. The IEC 62443 series addresses security for industrial automation and control systems. Part 4-1:2018, for example, sets requirements for a secure product development life cycle. ISO 13849-1:2023 also notes that security issues can affect safety functions. For ROS 2, SROS2 provides tooling to run on DDS-Security with keys, certificates and access control.

Where to start

Read What is a robot for the vocabulary, then the control hub to see where safety functions sit relative to motion control. The ROS 2 and middleware hub explains the graph that SROS2 protects. The Universal Robots and Franka platform pages cover two cobot makers. ISO’s robotics committee catalogue lists every published robotics standard.

Tools and platforms

Document or tool What it covers Publisher
ISO 12100:2010 General principles, risk assessment and risk reduction for machinery ISO
ISO 10218-1:2025 Safety of industrial robots as partly completed machinery ISO
ISO 10218-2:2025 Industrial robot applications and robot cells ISO
ISO/TS 15066:2016 Collaborative industrial robot systems ISO
ISO 13482:2014 Personal care robots, revision for service robots in progress ISO
ISO 13849-1:2023 Safety-related parts of control systems, performance levels ISO
IEC 62061:2021+AMD1:2024 Functional safety of safety-related control systems for machinery IEC
IEC 62443 series Security for industrial automation and control systems IEC
SROS2 ROS 2 on DDS-Security, Apache-2.0 ROS 2 project

Common mistakes

  • Certifying the robot and forgetting the cell. The assessment covers the tool, the part, the layout and the people.
  • Treating “cobot” as a safety rating. Collaboration is a property of the application.
  • Running a risk assessment once at design time and never after a change of tool, layout or program.
  • Implementing a safety function in ordinary application code. Safety functions run on safety-rated hardware and software.
  • Leaving controllers and ROS 2 graphs open on a shared network.
  • Quoting a standard from memory or a secondary summary. Buy the current edition and read it.

Going deeper

Recent papers

  • Runtime Safety and Emergency Stops on a Full-Size Humanoid: Deployment Lessons and a Case for Graduated Force Authority

    M. Conway, Y. Zhang, A. Devaraj et al. · arXiv cs.RO · digest

    The paper describes the safety layer the authors built into their open-source autonomy software for a Unitree H1-2 humanoid. The layer runs apart from the control software, checks every joint command against position, velocity and torque limits, and stops the robot when the state estimate goes stale or the hard-wired e-stop is pressed or unplugged. In push tests on the physical robot with a model-predictive lower-body controller, the robot never recovered its balance, the operator hit the e-stop every time, and because stopping removes balance control the robot fell. Joint limits chosen for manipulation were also exceeded in ordinary walking and triggered the same stop. The authors argue that next to an operator or a battery pack, falling may do more harm than whatever caused the stop and propose granting force authority in stages with a named responsible party, which is a position argued from the deployment rather than a tested method.