# ISO 10218 2025 changes: what robot cells must do differently

> ISO 10218-1:2025 and -2:2025 replace the 2011 editions. What changes for collaborative cells, functional safety, cybersecurity and Indian shop-floor compliance.

Canonical: https://roboticsengineering.in/iso-10218-2025-changes-india/ | Published: 2026-10-31 | Updated: 2026-10-31 | Author: Jignesh Panchal

**The 2025 revision of ISO 10218 is not a paperwork refresh. It removes the idea that a robot can be “collaborative” by model, and moves that judgement into the application — which means the risk assessment, not the datasheet, decides what safeguards a cell needs.**

## What the two documents are, and what changed in scope

Tenders still cite “ISO 10218” as one document. It has been two since 2011, and the split is what decides who is responsible for what.

|  | 2011 editions | 2025 editions |
| --- | --- | --- |
| Part 1 — the robot | Robot as a product: stopping functions, enabling device, protective stop | Same boundary, with functional safety expectations stated rather than implied |
| Part 2 — the system | Integration, layout, safeguards, operating modes, commissioning | Same boundary, with collaboration handled inside the system standard |
| Collaboration | Largely delegated to ISO/TS 15066 | Treated as a property of the application, assessed in the cell |
| Cybersecurity | Not addressed | Brought into scope, reflecting networked cells and remote access |

Both parts were published in February 2025. The 2011 editions remain in circulation in older drawings, older tenders and older integrator templates, which is why two cells built to different paperwork can both claim to follow “10218”.

Part 1 is the robot maker’s document. Part 2 is your integrator’s, and it is the one a plant engineer argues about, because layout, guarding, operating modes and commissioning all sit inside it. Our overview of [robot safety standards in India](https://roboticsengineering.in/robot-safety-standards-india/) places both parts next to the supporting standards they depend on.

## The change that matters most: collaboration is no longer a robot category

For fifteen years the industry used “collaborative robot” as a product class: buy this arm, and the collaboration question is answered. The 2025 editions separate the two ideas. Collaboration is a property of how a robot is used — the tool, the payload, the speed, the separation, the task — not a label the manufacturer can attach to the arm.

The practical consequence is blunt. A cell with a cobot arm, a sharp fixture and a 12 kg payload at full speed is not a collaborative application, whatever the brochure says. A cell with a conventional industrial arm, force limits set and verified, and a documented assessment can be. The datasheet tells you what the arm can do; the assessment tells you what this cell is allowed to do.

This is why [ISO/TS 15066 and cobot safety](https://roboticsengineering.in/cobot-safety-iso-ts-15066/) still matters after the revision: the Technical Specification holds the power and force limiting detail that applications are tested against, and the 2025 parts now assume that work is being done rather than outsourcing it to a separate document. A cobot cell without that assessment has a gap in exactly the place the revision tightened.

## Functional safety moved from implied to explicit

The 2011 text assumed that a safety function was designed properly. The 2025 editions are more direct about the control system behind the function: stopping, protective stop, enabling devices and safe speed each need a performance level, and that level has to be justified by architecture, component data and diagnostic coverage.

In practice this changes what an integrator hands over. “The gate is interlocked” stops being an answer. The answer is a named safety function, a target performance level, the circuit that achieves it, and a test that shows it. That is the same chain described in [robot safety relay wiring](https://roboticsengineering.in/robot-safety-relay-wiring-guide/), and the reason a single-channel switch with a two-channel symbol on the drawing is a finding rather than a detail.

Two consequences on a real floor:

- **Stop time becomes a record, not an assumption.** Every distance that depends on stopping time — curtain, scanner, gate position — inherits the measurement: robot, payload, tool, program, override, pose, instrument, date, person.
- **Change control becomes part of the safety case.** A new gripper or a speed increase is a change to the system, and the 2025 Parts are the benchmark a modified cell is judged against.

## Cybersecurity is now inside the standard

A robot cell is a networked machine. Program backups, remote support, MES links and vision systems all create paths into a controller that can move several hundred kilograms. The revision brings that into scope rather than leaving it to IT policy.

For an Indian plant the practical items are unglamorous: who can reach the controller, whether remote access is temporary and logged, whether program changes are versioned, and whether a USB stick is still the transfer mechanism. None of it requires a security programme; all of it fails an audit when it is missing from the handover file.

## What changes on an Indian shop floor

Very little hardware changes because a standard was revised. What changes is what you can be asked to produce.

1. **A risk assessment that names the tasks.** Loading, cleaning, jam clearing, maintenance and changeover, not just normal production. The assessment is where collaboration is accepted or refused, and where the 2025 framing bites.
2. **A safety function list with performance levels.** Gate, light curtain, scanner, e-stop, enabling device and safe speed, each with a target and a result.
3. **Stop-time records for every robot at the payload and speed in use.** A figure from the manufacturer at a different payload is an assumption, and assumptions are what the new text asks you to replace.
4. **Guarding that matches the hazard.** Fencing where parts are ejected, curtains where only reach is at stake, and interlocks that resist a magnet. The reasoning is in [robot guarding and fencing](https://roboticsengineering.in/robot-guarding-fencing-india/).
5. **A layout that reduces the need to enter.** The cheapest safety function is the access you designed out. [Robot cell layout design](https://roboticsengineering.in/robot-cell-layout-design/) is where that happens, before any steel is ordered.
6. **A validation report and an as-built pack.** Drawings, component list, interlock schematic, program archive and the test records, dated and signed.

None of this is new work for a well-run integration. It is new paperwork for an integration that was run on the assumption that the arm’s certification covered the cell.

## What the 2011 editions still mean for cells already running

A robot installed in 2014 was placed on the market under the standards of its time, and an unmodified installation is generally judged against that state of the art. You cannot re-certify the arm to a 2025 Part 1.

But the cell is a different question. Add a conveyor, change the gripper, raise the speed, change the product or move a guard, and the system has been modified — at which point the current editions are the benchmark. In practice the trigger is rarely the arm. It is the upgrade nobody wrote down.

For an older cell the sequence is the same as a retrofit: survey and as-built drawings, stop-time measurement at current payload, risk assessment of the cell as it stands, then a gap list. The document set that proves the result is the one in the [robot integration checklist](https://roboticsengineering.in/robot-integration-checklist/).

## How to tell which edition applies to your cell

1. **Find the date of the last risk assessment.** If it predates the last change to the cell, the paperwork is describing a machine that no longer exists.
2. **Ask for the performance level of each safety function.** If the answer is a component brand rather than a PL and a circuit, the design was not documented.
3. **Check whether stop time was measured or quoted.** Measured means a record with an instrument and a date.
4. **Check the interlocks for defeat.** A coded actuator and a discrepancy check are the difference between a switch and a safeguard.
5. **Check what the tender says.** A tender written before the revision and copied since will cite the 2011 text; the delivery should still be judged against the current edition.

Whether the 2025 editions have been adopted by BIS, and under which dual number, is a question for the BIS standards portal rather than for a supplier’s brochure.

## Frequently asked questions

**When were ISO 10218-1:2025 and ISO 10218-2:2025 published?**

Both parts were published in February 2025, the first revision since the 2011 editions. Part 1 covers the robot as a product; Part 2 covers the robot system and cell, including integration, safeguarding and commissioning.

**Does the 2025 revision make cobots safer by default?**

No. It removes the assumption that a cobot is collaborative by design and makes collaboration a property of the application. The same arm can be a collaborative application in one cell and a conventional guarded cell in another, depending on tool, payload, speed and task.

**Is ISO/TS 15066 obsolete now?**

The collaborative requirements now sit inside the 10218 series, but the 2016 Technical Specification is still widely cited in tenders and still holds the power and force limiting detail applications are tested against. Treat it as the detailed reference, not as a competing standard.

**Do I have to re-certify an existing robot cell to the 2025 editions?**

Not if the cell is unmodified. A change to the gripper, payload, speed, product, guarding or layout makes it a modified system, and the current editions become the benchmark for that reassessment.

**What is the single document an auditor asks for first?**

The risk assessment, dated and signed, covering the tasks people actually perform including cleaning and jam clearing. The validation report, the last stop-time record and the last functional test usually follow immediately.

**Does cybersecurity really affect a robot cell?**

Yes, once the controller is on a network. Remote support, MES links, program backups and USB transfers are all paths to a machine that can move heavy payloads. The controls are basic: named access, logged remote sessions, versioned programs and a known transfer method.

## Sources and further reading

- [ISO 10218-1:2025 — Robotics, safety requirements, Part 1](https://www.iso.org/standard/73933.html)
- [ISO 10218-2:2025 — Robot applications and robot cells](https://www.iso.org/standard/73934.html)
- [ISO/TS 15066:2016 — Collaborative robots](https://www.iso.org/standard/62996.html)
- [ISO 12100:2010 — Risk assessment and risk reduction](https://www.iso.org/standard/51528.html)
- [ISO 13857:2019 — Safety distances](https://www.iso.org/standard/69569.html)
- [ISO 14119:2024 — Interlocking devices](https://www.iso.org/standard/75942.html)
- [ISO 13855:2024 — Positioning of safeguards](https://www.iso.org/obp/ui/#iso:std:iso:13855:en)
- [HSE — PUWER 1998](https://www.hse.gov.uk/work-equipment-machinery/puwer.htm)
- [OSHA — Robotics](https://www.osha.gov/robotics)
- [BIS — Indian standards portal](https://standards.bis.gov.in/)