ISO/IEC DIS 42102 · stage 40.60 · 41 pages
Describe what your AI system does, before someone else does
Read from the real rules, not invented. Every answer points to the official rule behind it, so you can check any line yourself.
Read from the real rules167 official rules behind itCross-checked against 3 related standardsCheck any line yourself
The draft framework for characterizing AI system methods and capabilities gives you two axes and a shared vocabulary. This site gives you the public record, a free matrix check, the crosswalk to the AI standards you already run, a descriptor kit at USD 799 one off and a change watch at USD 49 a month.
The matrix, before a claim is made
| Sensory | Knowledge | Action | Comms | Emergent | |
|---|---|---|---|---|---|
| Traditional | |||||
| Symbolic | |||||
| Machine learning | |||||
| Hybrid |
Every claimed cell needs an evidence pointer and a date. Empty cells say why: not applicable, not yet evidenced, or outside scope.
The one minute answer
ISO/IEC DIS 42102 is a draft standard that gives organisations a common way to describe an AI system. It separates methods, which are how the system works, from capabilities, which are what it can do, and arranges them as a matrix. It is descriptive: no shall requirements, no controls, no certificate. ISO records stage 40.60, close of voting, and no publication date. This site gives you the public record, a free matrix check, a USD 799 kit and a USD 49 a month watch.
The useful work is not filling every cell. It is deciding which cells your organisation will claim, what evidence supports each claim, and who is accountable when a method changes. That is the difference between a taxonomy exercise and a description an auditor, a procurer and an engineer can all use.
What the public record proves, and what it does not
The draft exists and its structure is public
Seven clauses, forty one pages, committee ISO/IEC JTC 1/SC 42, ICS 35.240.01, edition 1. The public preview carries the scope, the introduction, the seven defined terms and the bibliography.
The stage and the dates
Stage 40.60, close of voting, effective 8 September 2026 on national body pages. The DIS ballot ran from 15 June to 7 August 2026.
The clause bodies and the annexes
The bodies of clauses 4 to 7 and the annex titles are not in the public preview. This site does not invent them and does not publish an Annex A page.
The methods dimension
Four families, and none of them is a maturity level. A rule engine is not a worse model; it is a different method with a different evidence story.
The capabilities dimension
Five groups, written as what a user can recognise. A capability claim should be testable by a stranger with the evidence you hold.
Two axes, one evidence rule
The matrix is the deliverable. We pair it with 167 controls exported from ISO/IEC 42001:2023, ISO/IEC 23894:2023 and ISO/IEC 42005:2025, so a claimed cell has somewhere to land.
Check one system in ten answers
Answer ten questions about one AI system. The check builds the method and capability grid your answers create, counts the cells that still need evidence, and returns at least three dated actions. It runs in your browser, sends nothing and returns scope incomplete rather than a pass when an answer is missing.
Kit and watch
The descriptor kit: the full register, the evidence list, the matrix template and the dated plan, as PDF, CSV and XLSX.
- All 33 units with the claim level stated
- A method and capability matrix template with the evidence rule
- The control crosswalk for 3 AI standards
- A dated plan you can work back from your review
The change watch: an email when the draft record, its stage or its dates change.
- Checks the ISO record and the public preview
- Emails the diff, not a newsletter
- Covers approval, publication and national adoption
- Cancel from the account page at any time
One click to checkout. No sign-in first. Prices in US dollars. The account is created from the Stripe email. Not legal advice; does not certify.
The rules this is built from
Read from the real rules, not invented. Every answer points to the clause behind it, so you can check any line yourself. The draft is read from its public preview; the evidence discipline comes from three AI standards used for the crosswalk.
ISO/IEC DIS 42102
The catalogue record, the scope, the introduction, the seven defined terms and the bibliography, read on 22 October 2026.
ISO/IEC 42001:2023
88 controls, 74 exported for this build, read for this build. The management system the buyer already runs.
ISO/IEC 23894 and 42005
54 risk controls and 39 impact assessment controls, exported for the crosswalk. Together with 42001 they make 167 rules behind the matrix.
How this compares with the alternatives
| Route | What you get | What it does not give you |
|---|---|---|
| The ISO catalogue and public preview | The authoritative stage, dates, scope, introduction and seven defined terms, free. | The clause bodies and annexes, which are not public, and any plan for your systems. |
| Buying the draft from ISO or a national body | The full descriptor wording, which you need before you finalise a template. | A crosswalk to the standards you already run, or an evidence list. |
| A general AI governance platform | A broad inventory, assessment and reporting workflow across many rules. | The 42102 structure at clause level, and a matrix built by a stated rule. |
| This site and the kit | The public record, a free matrix check, the control crosswalk, an evidence list and a dated plan. | The draft text itself, a certificate, or legal advice. |
Common questions
- What is ISO/IEC DIS 42102?
- It is a draft International Standard from ISO/IEC JTC 1/SC 42 that sets out a framework of descriptors for describing AI system methods and capabilities. It is descriptive, not a management system standard, and it sets no requirements or conformity test.
- When will it be published?
- ISO records stage 40.60, close of voting, with an effective date of 8 September 2026 on national member body pages. No publication date is published. This site uses 16 September 2027 as a planning estimate and labels it as one.
- What is a method and what is a capability?
- A method is how the system does something: rules, search, knowledge representation, machine learning or a combination. A capability is what it can do: process sensory input, process knowledge, control action, communicate or show emergent behaviour.
- Does the draft contain Annex A controls?
- No. The draft is not a requirements standard and the public preview does not publish annex titles. The introduction says informative annexes explain and illustrate the framework. This site does not invent an Annex A.
- What is in the kit?
- The full register of 33 units, the evidence list, the matrix template, the control crosswalk and a dated plan, as PDF, CSV and XLSX for USD 799 one off.
- Is this legal advice or a certificate?
- No. It is a reference and a planning aid. It does not certify, does not reproduce the draft and is not affiliated with ISO or IEC.
- What does the free check do?
- Ten answers build the method and capability matrix for one system, count the cells that still need evidence and return at least three dated actions. It runs in your browser and sends nothing.
Last updated: · Markdown version · Not legal advice; does not certify.