ISO 5338 establishes a set of processes for the lifecycle of AI systems based on machine learning and heuristic systems, with a special focus on the software part of the AI system.
The processes defined in ISO 5338 support the definition, implementation, validation, control, management, execution and improvement of AI systems throughout their lifecycle stages, following an engineering approach. These processes can be used in an organisation or project both when developing and acquiring AI systems.
ISO 33000 proposes a framework through which the maturity of the implementation of ISO 5338 processes in an organisation can be assessed, laying the foundations for the certification of conformance to these two standards.
Why is ISO 5338 important?
ISO/IEC 5338 is an essential standard for organisations developing and implementing artificial intelligence systems, as it establishes a process framework to support such development with considerations for the quality, reliability and transparency of these systems.
As AI becomes integrated into critical decision-making processes, ensuring the robustness and consistency of systems and their models is key to their effectiveness and acceptance. This standard provides principles and guidelines to help assess and improve the quality of AI systems, reducing bias, improving interpretability and ensuring alignment with ethical and regulatory requirements.
ISO 5338 provides a structured approach to adopting best practices in the development, verification and validation of AI systems, ensuring that they operate reliably and within expected parameters, thereby also promoting transparency and confidence in their use.
ISO 33000 establishes a framework for assessing the capability and maturity of processes in organisations. This standard allows organisations to identify strengths and areas for improvement, ensuring that their processes are efficient, repeatable and aligned with quality standards.
In conjunction with ISO 5338, organisations developing AI systems can implement an incremental framework for improving the maturity of their AI development lifecycle processes, which helps them to reduce risk, improve customer satisfaction and increase competitiveness.
Key principles of ISO 5338:
- Transparency and explainability – Provides principles to improve the interpretability of AI models, allowing users to understand how and why certain results are generated.
- Algorithmic bias reduction – Promotes practices to identify and mitigate bias in AI systems, ensuring fairer and more equitable decisions.
- Validation and verification – Defines methods for assessing the performance and safety of AI models, ensuring their alignment with expected requirements prior to implementation.
- AI Systems Reliability – Establishes guidelines to ensure that artificial intelligence models operate consistently and predictably in a variety of environments.
- Security and robustness – Encourages the development of AI models that are resistant to adversarial attacks and unexpected conditions, reducing deployment risks.
- Better integration with existing systems – Facilitates the implementation of AI into established infrastructures, ensuring compatibility and operational efficiency.
- Alignment with ethical and regulatory principles – Establishes a framework to ensure that AI systems comply with internationally accepted ethical standards and values.
Benefits of certification
- Improve the Quality and Maturity of AI Processes. It enables the evaluation and optimisation of AI system development processes, ensuring greater efficiency, repeatability and alignment with international standards.
- Assuring the Reliability and Robustness of Models. Provides a framework for validating and verifying AI systems, ensuring they operate consistently, minimising errors and operational risks.
- Reduce Bias and Promote Ethical Use of AI. Establishes practices to identify and mitigate biases in algorithms, promoting fairer decisions aligned with ethical principles.
- Comply with International Regulations and Standards. Ensures that AI system development processes comply with regulatory frameworks and transparency requirements, avoiding legal risks and improving user confidence.
- Optimise Project and Resource Management. It provides tools to measure the capacity of processes, allowing a better allocation of resources and a more efficient management of technological projects.
- Foster Innovation and Competitiveness. Help develop more advanced and reliable AI solutions, driving the adoption of best practices in the industry and improving the organisation's competitive advantage.
Key requirements and processes
ISO 5338 establishes a set of processes for the lifecycle of artificial intelligence (AI) systems. These processes reflect the following key points:
- Integrating security from the start: Emphasises the importance of incorporating security measures from the earliest stages of AI system development, recognising the unique vulnerabilities of these systems and promoting proactive threat management.
- Proactive risk management: Defines specific processes for continuous risk management, emphasising the evolving nature of AI threats and encouraging organisations to identify, assess and mitigate risks on an ongoing basis throughout the lifecycle.
- Importance of data management: Recognises the importance of quality, provenance, lineage and meticulous documentation of data in AI systems, which are fundamental to maintaining compliance with privacy regulations and protecting sensitive information.
- Continuous validation: Introduces the concept of continuous validation to ensure that the performance of AI systems remains robust and secure over time, allowing to detect and address security vulnerabilities, data deviations or conceptual changes that may compromise the integrity of the system.
Get certified with I2SC
At I2SC, we offer expert advice on the implementation and certification of ISO 33000 – ISO 5338. Our team will accompany you in each phase of the process to assess, improve and certify the maturity of your organisation through the quality of the AI system development lifecycle processes.
Contact us today and we will advise you on your path to ISO 33000 – ISO 5259 certification for your AI system development processes.
Are you interested in ISO 5338 certification?
EN 
ES