Borrador del Comité
ISO/IEC CD 25059
Software engineering — Systems and software Quality Requirements and Evaluation (SQuaRE) — Quality models for AI systems
Reference number
ISO/IEC CD 25059
Edition 2
Borrador del Comité
ISO/IEC CD 25059
88234
El comité está revisando un borrador.
Reemplazará ISO/IEC 25059:2023

What is ISO/IEC 25059?

ISO/IEC 25059 is an extension of the SQuaRE series that defines a quality model specifically for artificial intelligence (AI) systems. It adapts the principles of software and systems quality — as described in ISO/IEC 25010 — to address the unique properties of AI, such as probabilistic outcomes, learning behaviours, and reliance on data. This model helps organisations define, evaluate and measure AI quality consistently across products and tasks.

Why is ISO/IEC 25059 important?

Unlike conventional software, AI systems don’t always produce the same outputs for the same inputs, and their behaviour can evolve over time. Traditional quality models don’t fully capture these complexities. ISO/IEC 25059 closes that gap by offering a structured framework for describing and assessing AI system quality — from both the technical (product quality) and practical (quality in use) perspectives. It supports the development of trustworthy, high-performing AI by introducing new quality characteristics tailored to AI's unpredictable, data-driven nature.

Benefits

  • Establishes a clear framework for assessing AI system quality
  • Adapts conventional software quality models to AI-specific behaviour
  • Supports consistent specification, testing and evaluation of AI solutions
  • Helps identify gaps in quality requirements early in development
  • Provides terminology aligned with broader AI quality and trustworthiness standards

 

FAQ

AI developers, software engineers, quality assurance teams, and system evaluators seeking to define or measure quality in AI systems.

No. While it’s especially relevant for technical teams, it also supports stakeholders responsible for specifying requirements, testing outcomes, and ensuring the trustworthiness of AI solutions.

It accounts for things like learning behaviour, adaptation, data uncertainty, explainability, and real-world performance — all of which go beyond the scope of typical software quality models.

Informaciones generales

  •  : En desarrollo
    : Estudio de CD iniciado [30.20]
  •  : 2
  • ISO/IEC JTC 1/SC 42
  • RSS actualizaciones

¿Tiene alguna duda?

Consulte nuestras Ayuda y asistencia