Tidewell Robotics

The standard your robot is not covered by

ISO 10218-1:2025 excludes medical robots, healthcare robots and service robots the public can access, and names no standard that would reach them. ISO 3691-4:2023 carves out the additional hazards of trucks intended to operate in public zones; ISO/CD 25785-1 defines its industrial environment as a workplace where the public is excluded or restricted from access; both name ISO 13482 as the standard for what they exclude. Three standards push a legged robot in a hospital out, two of them pointing at a fourth whose in-force edition covers three personal care robot types it is not one of, while the edition written for service robots is an unpublished draft. That is why a safety file for that machine is built from ISO 12100 upward, not claimed.

Insight · 12 August 2026 · Updated 11 September 2026 · 13 min read · Tidewell Article Crew, edited by Timothy Mo

A guide to commercial cleaning robot safety, dated 12 March 2024 and still live when we fetched it on 11 September 2026, opens its standards section like this:

"ISO 10218 is one of the most notable standards, which addresses the safety requirements for industrial robots. It covers aspects such as robot design, system integration, and safeguarding measures to protect operators and bystanders from potential hazards associated with robotic operations."

The guide is published by Gausium. ISO 10218-1:2025's catalogue record says the standard "is not applicable to the following uses and products: … medical robots; healthcare robots; … service robots, which provide a service to a person and as such where the public can have access; consumer products, as this is household use to which the public can have access…" A floor machine working a mall concourse or a hospital corridor is a service robot the public can reach. The standard the guide leads with is the one that writes the guide's subject out.

Two further designations appear in the same guide: "ISO 18650 is a fundamental standard focusing on the safety requirements for household and similar electrical appliances, including those used in commercial settings", and "ISO 18489 is another essential standard that addresses the safety and performance of commercial robotic equipment, including cleaning robots." Neither could be found in ISO's catalogue this pass, and how the lookup failed matters: ISO's catalogue search is a JavaScript application and www.iso.org returned 403 to every request we made on 11 September 2026, so we could read any record we had an identifier for but could not run a search. Not found is not the same as not existing. What can be checked without a search is that household and similar electrical appliance safety is the IEC 60335 family, an IEC series rather than an ISO one, so the attribution is to the wrong body at minimum [inference].

The guide's reference list then cites "ISO 10218-1:2011(en) Robots and robotic devices — Safety requirements for industrial robots." and links to the record for the edition published 1 July 2011 and withdrawn at stage 95.99 on 5 February 2025.

A correct claim looks different. Panasonic's HOSPI hospital delivery robot is ISO 13482 certified, and Yujin Robot's GoCart was Korea's first ISO 13482-certified AMR [background] — service robots certified to the service-robot standard.

The working notes behind this article carried two scope phrasings that turn out not to be in the texts at all; reading the catalogue records line by line is what catches that.

The argument: ISO 10218-1:2025, ISO 3691-4:2023 and ISO/CD 25785-1 each exclude a legged robot working around the public in a hospital. Two of them name ISO 13482 in their own text as where such a machine belongs. ISO 10218-1 names no destination at all; it excludes and stops. And ISO 13482's in-force edition covers three personal care robot types that machine is not one of, while the edition retitled for service robots is an unpublished draft. The chain ends nowhere, which is why a safety file for such a machine is built from ISO 12100 risk assessment upward, citing clauses as evidence rather than claiming conformity to one.

The scopes, read out loud

We read every stage, edition and exclusion below off ISO's committee catalogue on 11 September 2026. We did not buy the texts, so these are the published catalogue records, not numbered scope clauses.

ISO 10218-1:2025, Robotics — Safety requirements — Part 1: Industrial robots, Edition 3, ISO/TC 299, reached stage 60.60 on 5 February 2025 — the day the 2011 edition went to 95.99. Its record carries two lists, and the difference between them is where the confusion starts. One is the products list quoted above; the other is hazards. The document "does not cover the hazards related to: … mobility when robots or manipulators are fixed to or part of driverless industrial trucks; mobility when robots or manipulators are fixed to or part of mobile platforms; … when the public, all ages or non-working adults have access (e.g. service robots, consumer products)." So ISO 10218-1 does not exclude mobile robots as a product. It excludes medical robots, healthcare robots and service robots the public can access, and covers none of the hazards mobility creates. That is how a vendor can say its arm meets ISO 10218-1 and be technically right, while the standard is silent on the walking and its products list excludes the machine the arm is bolted to [inference].

ISO 3691-4:2023, Industrial trucks — Safety requirements and verification — Part 4: Driverless industrial trucks and their systems, Edition 2, reached 60.60 on 12 June 2023 and has sat at 90.92 since 6 December 2023. Edition 1 published 17 February 2020 and was withdrawn the day Edition 2 published; neither record carries an Amd 1:2023, so the shape people write, the 2020 edition as amended in 2023, does not exist. Its scope covers driverless industrial trucks. Its exclusions read: "It does not apply to additional hazards that can occur: … from trucks intended to operate in public zones (see in particular ISO 13482:2014); … during operation with specific hygienic requirements…" Note where hygiene sits: a hazard carve-out, not an environment carve-out, and the words hygienic environment are not in the text. Edition 3 is at DIS, stage 40.20, ballot initiated 17 June 2026 for twelve weeks; stage 40.60 was blank on 11 September 2026, so we do not say the ballot has closed.

ISO/CD 25785-1, Robotics — Safety requirements for dynamically stable industrial mobile robots (legged, wheeled, or other forms of locomotion) — Part 1: Robots, is at stage 30.60, close of comment period, dated 8 July 2026, everything from 30.99 onward blank. It is the document being drafted for quadrupeds, bipeds and wheeled balancing machines. It defines the space it governs — "For the purpose of this document, an 'industrial environment' is a workplace where the public is excluded or restricted from access." — and says where everything else goes: "Note 2: Safety requirements for mobile robots used in non-industrial environments can be found in other standards (e.g., ISO 13482)."

The out-of-scope list goes further: "The following robots are out of scope of this document: … commercial and industrial floor-cleaning robots; and robots when used in non-industrial environments." The standard drafted for legged and balancing machines writes commercial floor cleaners out by name, while the category's safety literature writes them into the industrial-robot standard.

ISO 13482 is two documents. In force is ISO 13482:2014, published 3 February 2014, at 90.92 since 23 July 2021 and scoped by robot type rather than environment: it "specifies requirements and guidelines for the inherently safe design, protective measures, and information for use of personal care robots, in particular the following three types of personal care robots: mobile servant robot; physical assistant robot; person carrier robot." The second edition, ISO/FDIS 13482, retitled Robotics — Safety requirements for service robots, reached stage 50.00 on 24 July 2025, and stages 50.20, 50.60 and 60.60 are blank: there is no publication date on the record, and anyone quoting one is quoting something else. ISO's published abstract says it "is not intended to cover robots used in industrial applications and medical applications."

Two more documents belong in the count and are rarely named. ISO/WD 26058-1, at stage 20.60, comments closing 18 December 2025: "The following automated mobile machines are out of scope of this document: … Commercial and industrial floor-cleaning robots; and Mobile machines when used in non-industrial environments." And ISO/TS 15066:2016, cited everywhere for collaborative contact forces, whose abstract says: "It does not apply to non-industrial robots, although the safety principles presented can be useful to other areas of robotics."

The chain, and where it ends

The exclusions are not parallel. These documents hand the machine to each other, in writing, and the last hand is empty. ISO 3691-4:2023 carves out hazards "from trucks intended to operate in public zones (see in particular ISO 13482:2014)". ISO/CD 25785-1's Note 2 sends non-industrial mobile robots to "other standards (e.g., ISO 13482)". ISO 13482:2014 in turn excludes "industrial robots, which are covered in ISO 10218", and ISO 10218-1:2025 excludes service robots the public can access. Two of those referrals point at ISO 13482. The third is ISO 13482's own, and it points back. ISO 10218-1 points nowhere. What that leaves is a closed loop with a single exit: at the exit, the in-force text offers three personal care types, and the text that would fit is a draft.

  1. Three standards, each putting the machine outside

    ISO 10218-1:2025Published, stage 60.60, 5 February 2025. "This document is not applicable to the following uses and products: … medical robots; healthcare robots; … service robots, which provide a service to a person and as such where the public can have access…"
    ISO 3691-4:2023Published, stage 60.60, 12 June 2023. "It does not apply to additional hazards that can occur: … from trucks intended to operate in public zones (see in particular ISO 13482:2014)…"
    ISO/CD 25785-1Draft, stage 30.60, comment period closed 8 July 2026. "Note 2: Safety requirements for mobile robots used in non-industrial environments can be found in other standards (e.g., ISO 13482)."
  2. ISO 13482Two of the clauses above name it. ISO 10218-1 names no destination at all — its products list simply excludes the machine.
  3. And ISO 13482 is two documents

    ISO 13482:2014 — in forceStage 90.92 since 23 July 2021. "…requirements and guidelines for the inherently safe design, protective measures, and information for use of personal care robots, in particular the following three types of personal care robots: mobile servant robot; physical assistant robot; person carrier robot."
    ISO/FDIS 13482 — not publishedStage 50.00 since 24 July 2025; stages 50.20, 50.60 and 60.60 are blank on the record, so there is no publication date to quote. Retitled Robotics — Safety requirements for service robots. The published ISO abstract: "This standard is not intended to cover robots used in industrial applications and medical applications."
  4. Nothing published downstreamNo published International Standard specifies safety requirements for a dynamically stable mobile manipulator working where the public can go. This last step is our reading of the five records above, not a sentence in any of them [inference].

Two of the three name ISO 13482 in their own text. The third names nowhere.

  • Published International Standard, in force
  • Where the referrals converge
  • Draft: registered, no publication date
  • Ours, not a standard's sentence [inference]
The edges are the standards' own words; the dead end is ours to draw. Every clause here is transcribed from ISO's catalogue records at committee.iso.org, read 11 September 2026 — records 73933, 83545, 91469, 53820 and 83498 — so the chain can be checked one record at a time.

The composite claim is ours, not ISO's, so it runs as a test a reader can repeat [inference]. The machine: a legged, dynamically stable mobile manipulator, working in a hospital, in spaces the public can enter.

StepStandardThe clauseDoes it reach the machine?
1ISO 10218-1:2025"… medical robots; healthcare robots; … service robots, which provide a service to a person and as such where the public can have access"No. Three independent bullets each exclude it, and the hazards of a robot "fixed to or part of mobile platforms" are outside what it covers, so even the arm-only reading leaves the walking uncovered.
2ISO 3691-4:2023"… from trucks intended to operate in public zones (see in particular ISO 13482:2014); … during operation with specific hygienic requirements…"No. A legged machine is not a powered truck; grant the truck reading anyway and the public-zone carve-out removes it, the hygienic carve-out the cleaning duty cycle.
3ISO/CD 25785-1"For the purpose of this document, an 'industrial environment' is a workplace where the public is excluded or restricted from access." Out of scope: "robots when used in non-industrial environments"No. A ward or corridor the public can enter is not that workplace, and if the machine cleans floors the floor-cleaning bullet catches it independently.
4ISO 134822014, in force at 90.92: "mobile servant robot; physical assistant robot; person carrier robot". FDIS, stage 50.00: "not intended to cover robots used in industrial applications and medical applications"The only one that plausibly reaches it, and it is not available. The in-force text forces the machine into one of three personal care types it does not match; under the FDIS it fits, and the FDIS is unpublished.

Three attacks on it are worth stating, because an argument that hides its weakest point is marketing. A hospital corridor is restricted access, so it is an industrial environment under ISO/CD 25785-1. Visitors, outpatients and patients are members of the public, and ISO 3691-4 carves out public zones for the same reason; two documents drawing the same line is not a drafting accident. ISO 10218-1 still covers the manipulator. Partly, and that is the point: partial coverage of a subsystem is evidence you cite, not conformity you claim. ISO 13482:2014 already covers it as a mobile servant robot. Arguable for a delivery robot, weak for a manipulator, and the second edition was retitled for service robots because the 2014 type list stopped describing the field [inference].

What you build instead

When no type-C standard's scope reaches a machine, the type-A standard still does, because it is not scoped to a machine at all. ISO 12100:2010, Safety of machinery — General principles for design — Risk assessment and risk reduction, reached 60.60 on 20 October 2010 and says in its abstract that it "is also intended to be used as a basis for the preparation of type-B or type-C safety standards." It belongs to ISO/TC 199, not the TC 299 that owns every robot standard above, and is itself at 90.92 since 1 December 2023, with ISO/DIS 12100.3 at stage 40.20 since a ballot opened on 13 December 2024 — the foundation is moving too.

So you run the ISO 12100 risk assessment and, for each hazard it identifies, cite whichever type-B or type-C clause carries the best-evidenced requirement as evidence that the measure is adequate. These documents already see themselves that way: ISO 3691-4:2023 defines its coverage by reference to clause 5.4 of ISO 12100:2010, and ISO/WD 26058-1 calls itself a Type C standard. What you cannot do is put a standard number on a certificate: conformity assessment runs against a document whose scope covers the product, and none does.

  1. ISO 12100:2010 — the type-A baseStage 90.92 since 1 December 2023, ISO/TC 199 rather than the TC 299 that owns every robot standard above, with ISO/DIS 12100.3 at stage 40.20. Its abstract: "ISO 12100:2010 is also intended to be used as a basis for the preparation of type-B or type-C safety standards." It is scoped to no machine, so no machine falls outside it.
  2. The risk assessment, hazard by hazardThese documents already see themselves this way: ISO 3691-4:2023 defines its coverage by reference to clause 5.4 of ISO 12100:2010, and ISO/WD 26058-1 calls itself a Type C standard.
  3. Per hazard, a clause cited upward as evidence

    A clause from a standard whose scope excludes the machineISO 10218-1:2025 does cover the manipulator, partly. Partial coverage of a subsystem is evidence you cite, not conformity you claim.
    A clause from a standard written for a neighbouring machineWhichever type-B or type-C clause carries the best-evidenced requirement for that hazard.

    Evidence that the measure is adequate, never a conformity claim — the scope of each of these documents excludes the machine.

  4. One safety fileWhich clauses of which standards it cites as evidence — the first half of the only question that separates a real file from a number.
  5. An assessor signsAnd who assessed them is the second half. What cannot sit here is a certificate naming a standard: conformity assessment runs against a document whose scope covers the product, and none does.

Every step here is a step in building a file, not a step toward a certificate.

  • The type-A base, scoped to no machine
  • A clause cited as evidence for one hazard
  • Where a person signs
The general shape of a safety file for a machine no product standard's scope reaches — not Tidewell's file. Note what the shape cannot produce: a certificate naming a standard, because conformity assessment runs against a document whose scope covers the product.

Three corrections come with this territory, each the same class of error as the opening exhibit.

ISO/TS 15066 was not withdrawn. Its content went into the 2025 ISO 10218 series — A3, which holds the ISO/TC 299 WG3 secretariat, says of Part 1 that "Where appropriate, ISO/TS 15066:2016 on the safety of collaborative robot applications was added to the ISO 10218 series", and of Part 2 that "Safety requirements for collaborative applications (formerly, the content of ISO/TS 15066) have also been incorporated." The document itself sits at stage 90.92, confirmed in 2022, with no withdrawal stage. Right about the content and wrong about the document is how that claim usually arrives.

ANSI/A3 R15.06-2025 inherits the exclusions. It "is the U.S. national adoption of ISO 10218 Parts 1 and 2 and is a revision of ANSI/RIA R15.06-2012", in A3's own words; a national adoption carries the adopted scope and therefore the adopted exclusions [inference]. Parts 1 and 2 have been available from A3 since 9 September 2025. The US mobile series does have a name for this machine — "IMR Type C: a mobile manipulator … mounted to a mobile platform" — and draws the same boundary: "Industrial robots operate in structured or semi-structured workplaces that are only open to trained personnel, not the public."

And nobody can say OSHA recognises R15.08. Its Technical Manual, Section IV, Chapter 4, read 11 September 2026, states that "The safety standard for robots is Part-1 and Part-2 of American National Standards Institute (ANSI)/Robotic Industries Association (RIA) R15.06-2012", and that "ANSI/RIA R15.06-2012 will be updated following the revision of ISO 10218-1 and ISO 10218-2, which is in progress" — present tense, nineteen months after ISO 10218:2025 published. A string search of the chapter returns zero occurrences of "R15.08". It describes the machine — "IMRs can navigate autonomously within their operating environment to reach specified locations, and are designed to automate transport tasks." — and then cites no mobile-robot standard for it.

What a buyer can ask for

The one document here that does not carve the public out governs the organisation rather than the machine. ISO 31101:2023, Robotics — Application services provided by service robots — Safety management systems requirements, published at 60.60 on 10 November 2023, is scoped to a provider operating "in unstructured human spaces with trained and untrained persons (e.g. giving directions for visitors in airport or shopping mall, carrying goods to patients in hospital, delivering food to customers in restaurant.)" — the environment every product standard above excludes. It says of itself that "This document is not intended to be used as a product safety standard."

It is written in management-system form and anticipates an organisation that will "demonstrate conformity with this document", the structural precondition for a third-party audit [inference] — though no accredited certification body for it was identified this pass, so name it in a tender and ask who would assess it.

Our own position is published and unflattering: the Body page states that we hold no ISO 13482, ISO 3691-4 or ISO 10218 certificate, and the safety page carries a per-product, per-standard grid in which no row claims a certificate — every row reads in-design or not-started, save one that reads not applicable. That is the state of a category, not a confession peculiar to us; these machines are uncovered whoever builds them. The instrument that will demand harmonised standards for machinery of this class is coming; that is the next article.

Until then, the question a procurement officer can put in a document: which clauses of which standards does your safety file cite as evidence, and who assessed them? A vendor with a real file answers in a paragraph. A vendor citing a standard whose scope excludes the machine answers with a number.