Robot Safety Standards: ISO 10218, ISO/TS 15066 and What They Actually Require

What does ISO 10218 actually require now?

ISO 10218 is the core international safety standard for industrial robots, and as of 2025 it looks meaningfully different from the version most engineers grew up with. ISO 10218-1:2025 (covering the robot itself) and ISO 10218-2:2025 (covering the robot cell and application) were published in February 2025 and took effect on April 1, 2025, replacing the 2011 editions. This is the first major overhaul of the standard in fourteen years, the product of nearly eight years of committee work. In practical terms, the new standard requires manufacturers and integrators to: classify every robot as Class I or Class II based on physical parameters, treat collaborative-application safety as a normative (not optional) part of the robot standard rather than a separate technical specification, and — for the first time — address cybersecurity as an explicit safety requirement rather than an afterthought. Anyone still designing a cell against the 2011 text is working from a document that no longer reflects current practice, even if legal enforcement dates in specific jurisdictions are still catching up.

The rest of this article unpacks each of those pieces, and clears up the most common confusion in the industry: what actually happened to ISO/TS 15066, the specification most people associate with “collaborative robot safety.”

What changed between the 2011 and 2025 editions?

The 2011 edition of ISO 10218 was written primarily around traditional industrial robots operating behind fences, with collaborative operation treated as an edge case. The 2025 revision assumes collaborative and shared workspaces are a normal part of industrial robot deployment, not a niche. According to The Robot Report and the Association for Advancing Automation (A3), the revision process took close to eight years and represents the first substantial rewrite since 2011 — not a minor amendment cycle.

Three changes stand out as functionally significant rather than editorial:

• A formal two-tier robot classification (Class I / Class II) replacing informal, application-specific risk judgments.
• Direct incorporation of collaborative-application requirements previously published separately as ISO/TS 15066:2016.
• A new, standalone cybersecurity requirements section for industrial robots.

Each of these has direct consequences for how a machine builder documents conformity and how an integrator designs a cell.

How does the new Class I / Class II classification work?

ISO 10218-1:2025 introduces a binary classification that determines which safety and control requirements apply to a given robot. Class I covers robots with a manipulator mass of 10 kg or less, a maximum force output of 50 N or less, and a maximum speed of 250 mm/s or less; these typically only need to meet Performance Level b (PL b) under ISO 13849-1. Any robot that exceeds even one of those three thresholds — mass, force, or speed — is automatically Class II, which carries the fuller set of control-reliability and risk-reduction requirements associated with larger, faster, or higher-force industrial machines.

This is a meaningful simplification for the small-payload, tabletop and desktop robot segment: it gives manufacturers of very small collaborative arms a clear, threshold-based route to a lighter compliance burden, rather than a case-by-case risk assessment against the full industrial standard. For anyone specifying a robot for a shared workspace, checking whether a given SKU falls under Class I or Class II is now one of the first steps in a safety file, not a downstream detail.

What happened to ISO/TS 15066 — is it gone?

This is the question that generates the most confusion, so it’s worth being precise. ISO/TS 15066:2016 was originally published as a standalone technical specification supplementing ISO 10218, specifically addressing collaborative robot applications — the “power and force limiting” (PFL) mode that lets a robot and a person share a workspace without a fence. With the 2025 revision, the requirements that used to live in ISO/TS 15066 have been folded directly and normatively into ISO 10218-2:2025. ISO itself, along with ANSI’s summary of the revision, describes this as full integration rather than a parallel update — ISO/TS 15066 no longer functions as an independent, separately-cited document.

Practically, this means engineering teams should stop treating “ISO 10218 compliance” and “ISO/TS 15066 compliance” as two separate boxes to check for a collaborative cell. The biomechanical limits and collaborative-operation logic that ISO/TS 15066 defined are now part of the same normative text as the rest of the robot and cell requirements, so a single conformity assessment against the 2025 editions covers what used to require citing both documents.

What did ISO/TS 15066 actually specify about contact limits?

Even as it gets absorbed into ISO 10218, the technical content of ISO/TS 15066 is still the scientific basis for collaborative robot safety, so it’s worth understanding what it actually measures. The specification’s power-and-force-limiting approach is built on a body model covering 29 distinct body areas grouped into 12 regions — from fingertips to the skull to the abdomen. Annex A of the specification sets maximum permissible contact pressure and force values for each of these areas, derived from pain-threshold research conducted on 100 test subjects at the University of Mainz, with the data further adjusted to account for differences between sexes.

This is the reason collaborative robot speed and force limits vary so much by body region in a proper risk assessment: a contact event near the skull has a dramatically lower force ceiling than one on the forearm, and any PFL-mode robot deployment has to be tuned against this region-by-region table rather than a single blanket speed limit.

What’s new in the cybersecurity requirements?

For the first time, ISO 10218-1:2025 contains a dedicated section addressing cybersecurity for industrial robots, rather than leaving it to be inferred from general machinery safety principles. Per coverage from The Robot Report and ANSI’s technical summary, this section requires manufacturers to perform a cyber-threat assessment as part of the overall risk assessment, secure communication interfaces against unauthorized access, implement user authentication mechanisms, and — where the risk assessment indicates a credible threat — deploy encrypted communication protocols.

This lands the robot safety standard squarely on territory that used to be handled separately (if at all) by IT or OT security teams. For integrators building connected or fleet-managed robot cells, it means a cybersecurity risk assessment is now an expected part of the same documentation package as the mechanical and functional-safety risk assessment, not a bolt-on. That connects directly to the broader operational discipline of keeping a fleet of networked robots patched, monitored, and access-controlled, a topic covered in more depth in our explainer on running a robot fleet in production.

ISO 10218:2011 vs. ISO 10218:2025 — quick comparison

Aspect	2011 edition	2025 edition
Collaborative operation requirements	Referenced separately via ISO/TS 15066:2016	Integrated normatively into ISO 10218-2
Robot classification	No formal mass/force/speed tiering	Class I (≤10 kg, ≤50 N, ≤250 mm/s, typically PL b) vs. Class II
Cybersecurity	Not addressed	Dedicated requirements: threat assessment, secured interfaces, authentication, encryption where warranted
Scope assumption	Fenced, non-collaborative operation as default	Collaborative and shared workspaces treated as standard practice
Age of standard at revision	In force since 2011	First major overhaul in roughly 14 years, ~8 years in development

When do companies actually need to comply?

This is where sources genuinely diverge, and it’s worth stating that plainly rather than picking a single number. The standards themselves are already in effect: publication was in February 2025 and the effective date was April 1, 2025. But “in effect” as an ISO document and “legally mandatory for CE marking or procurement” are two different things. Industry coverage from the Association for Advancing Automation describes a proposed 24-month transition period before references to the old 2011 editions are fully withdrawn from certification practice, but whether the European Commission will formally accept that transition timeline in the Official Journal of the EU is not yet settled. Separate reporting on supplier readiness suggests full legal implementation in Europe — tied to the broader Machinery Regulation — is expected around 2027, though that date is an estimate rather than a confirmed regulatory deadline. If your organization is building a compliance roadmap, the safe planning assumption is: design to the 2025 requirements now, but confirm the specific transition deadline with your notified body or certification partner rather than assuming a fixed calendar date, since that detail is still in flux. For a wider view of how this fits into machinery regulation generally, see our overview of how the EU AI Act interacts with machinery rules for robots.

What should integrators and buyers do differently right now?

For teams actively specifying or deploying robots, three practical steps follow from the revision. First, when evaluating a robot for a shared or collaborative application, ask the manufacturer directly whether it has been assessed as Class I or Class II under ISO 10218-1:2025, and don’t assume a small robot automatically qualifies for the lighter tier — the mass, force, and speed thresholds all have to hold simultaneously. Second, treat any documentation that still separately cites ISO/TS 15066:2016 as a signal to ask whether the safety file has been updated to the integrated 2025 text, since a fragmented citation can indicate stale paperwork rather than a genuine compliance gap. Third, add a cybersecurity threat assessment to the standard risk-assessment checklist for any networked or fleet-managed robot cell, since this is now an explicit expectation rather than a best practice left to individual judgment.

None of this replaces a proper risk assessment conducted with a qualified safety engineer, and the transition-period uncertainty around CE enforcement means legal teams should stay close to their notified body’s specific guidance rather than working off press coverage alone. But understanding what actually changed — classification, integration of collaborative requirements, and cybersecurity — is the difference between reading a standard and applying it correctly. For the bigger picture of how safety regulation shapes robot deployment economics across the industry, our market outlook for physical AI covers how compliance costs and timelines factor into procurement decisions, and our glossary has quick definitions for terms like Performance Level and power-and-force limiting if you need a refresher while reading a safety file.

Robot safety standards move slowly by design, but the 2025 revision to ISO 10218 is the biggest structural change the industrial robotics field has seen in over a decade — and given the pace at which collaborative and networked deployments are growing, it’s unlikely to be the last one this side of 2030.