Second edition at FDIS, stage 50.00, 23 July 2026/ Planning date 15 February 2027, an estimate/ ISO/IEC 25059:2023 applies today
25059ISO 25059 Revision CheckerThe AI quality model, clause by clause, and the edition that follows it
ANSWERS

Five plain answers about ISO 25059

Short, direct answers to the questions a governance lead is asked first. Each has its own page with the reasoning, the source and the limit.

01

What is ISO/IEC 25059?

ISO/IEC 25059 is the SQuaRE quality model for AI systems. It extends the general product quality and quality-in-use models with AI-specific characteristics: user controllability, functional adaptability, functional correctness, robustness, transparency and intervenability. It is a model for specifying, measuring and evaluating quality, not a certification standard.

02

What changes in the ISO 25059 revision?

The public draft lists four changes: alignment with ISO/IEC 25010:2023, alignment with ISO/IEC 25019:2023, an AI service quality model based on ISO/IEC TS 25011:2017, and environmental sustainability as a sub-characteristic of performance efficiency. The title becomes plural because the document now holds three models.

03

Is ISO/IEC 25059 certifiable?

No. ISO/IEC 25059 defines a quality model. It sets no requirements an organisation can be certified against, and there is no accreditation or transition scheme for it. A buyer can adopt the model in a quality plan or a contract and an assessor can examine the evidence, but that is not a certificate.

04

ISO 25059 vs ISO 25010: what is the difference?

ISO/IEC 25010 defines the general product quality model and, in its 2023 edition, only the product model. ISO/IEC 25059 is an extension for AI systems: it adds and changes the sub-characteristics where AI behaviour needs it and inherits the rest. Quality in use now sits in ISO/IEC 25019:2023, which the second edition of 25059 aligns to.

05

How do you write an AI quality requirement?

Name the characteristic, state the task and the context, give a measure and a method, set a threshold, state the conditions the result holds under, name the evidence and give an owner. A requirement that repeats a definition without a measure cannot be evaluated, and a single accuracy figure is not a correctness requirement.