THE 25 RECORDS

ISO 25059, one page per clause and annex

What each model element is, what the 2023 edition said, what the public draft changes, the evidence a reviewer asks for, the ways it commonly fails and what to do before publication. Every page names its clause and its source. The change line is explicit about what the public record does not show.

Scope, references and terms

Clause	Record	Status	What it asks, in one line
1	Scope	Carried forward	The document sets out quality models that cover AI systems and the services built around them.
2	Normative references	Amended	The model stands on a small set of normative references: the product quality model, the quality-in-use model, the AI concepts and terminology standard, and the machine learning framework standard.
3.1	General terms	Amended	The general terms fix what a measure is, what a software quality measure is, what a risk treatment measure is, and what transparency means.
3.2	Product quality terms	Amended	The product quality terms define the characteristics the model adds or changes: user controllability, functional adaptability, functional correctness, intervenability and robustness.
3.3	Quality-in-use terms	Amended	The quality-in-use terms define the characteristics seen when the system is used in its context.
3.4	AI service quality terms	New in the second edition	New in the second edition.
4	Abbreviated terms	Carried forward	The clause lists two abbreviations: AI for artificial intelligence and ML for machine learning.

Product quality model

Clause	Record	Status	What it asks, in one line
5.1	General	Carried forward	The product quality model is a modified version of the general product quality model.
5.2	User controllability	Moved in the second edition	User controllability describes how far a user is able to step into how the system behaves, and to do so in time to matter.
5.3	Functional adaptability	Carried forward	Functional adaptability describes how far the system can adapt to a changing environment, including by taking in fresh training data, live production data and what earlier actions produced.
5.4	Functional correctness	Amended	The characteristic describes how far the system gives the correct results with the needed precision.
5.5	Robustness	Carried forward	Robustness describes how far the system keeps its level of correctness under difficult conditions: unseen, biased, adversarial or invalid inputs, outside interference, environmental conditions, and the ordinary expectation that the system works as its developers intended.
5.6	Transparency	Carried forward	Transparency describes how far useful information about the system reaches the people who need it: what it does, how it is put together, its interfaces, the models, the data used for training and evaluation, benchmark results, logs and the management practices behind it.
5.7	Intervenability	Moved in the second edition	Intervenability describes how far an operator can step in to prevent harm or hazard.
5.8	Environmental sustainability	New in the second edition	New in the second edition, under performance efficiency.

Quality-in-use model

Clause	Record	Status	What it asks, in one line
6.1	General	Amended	The quality-in-use model looks at the system in its context of use rather than at the product alone.
6.2	Transparency	Carried forward	Transparency in use is about whether the people affected and the people operating the system get information they can act on: what the system is for, where it should not be used, what its limits are and how its outputs should be read.
6.3	Societal and ethical risk mitigation	Amended	The characteristic describes how far the system reduces risk of a societal or ethical kind in its context of use.

AI service quality model

Clause	Record	Status	What it asks, in one line
7.1	General	New in the second edition	New in the second edition.
7.2	Traceability	New in the second edition	Traceability describes how far what the service produces can be connected back to what the user needed, or forward from that need.
7.3	Service adaptability	New in the second edition	Service adaptability describes how far the service can be adapted to the situation: its functions, who is responsible, the skill of the staff, the way it communicates with users and the instruments available.
7.4	Customizability	New in the second edition	Customizability describes how far the AI service can be changed at the user's request.

Clause numbers and titles are the standard's own, because a reader has to find the clause. Everything else is our wording. The three annexes are informative except where a page says otherwise; they have their own index. The change table shows the six new elements, two moves and seven amendments in one view.

THE MODEL

The rules this is built from

Every line names its source. The 25 records are written from the published ISO/IEC 25059:2023, in our words, with the clause cited. The change layer is read from the public draft of the second edition and the ISO stage record. Where the public record is silent, the page says so.

25 RECORDS

Clause by clause, and the annexes

Each page states what the model element is, what the 2023 edition said, what the public draft changes, the evidence a reviewer asks for and the common gaps.

Open the record

6 NEW, 2 MOVED, 7 AMENDED

The change, labelled

The second edition adds a service model and environmental sustainability, and moves two characteristics to better parents. Each page says which and why.

What changes

ONE STAGE RECORD

No invented date

The FDIS is at 50.00. ISO records no publication date. The planning date is an estimate and is labelled wherever it appears.

Read the revision record

The ISO/IEC 25059 text is not reproduced. The FDIS text is not public and is not reproduced. The framework this site reads is linked on the standards page.