Safety-Related Systems in Road Vehicles with Artificial Intelligence Are Addressed in ISO/PAS 8800:2024
In December 2024, the new publicly available specification ISO/PAS 8800:2024 Road Vehicles — Safety and Artificial Intelligence was released. This document is an important milestone in the development of standards for artificial intelligence (AI) and machine learning (ML) in the context of safety-critical automotive systems. ISO/PAS 8800 addresses safety-related properties and risk factors impacting the insufficient performance and malfunctioning behavior of Artificial Intelligence (AI) within a road vehicle context. These specifications were developed by industry and academic experts from 17 countries under the Road Vehicles Safety and Artificial Intelligence AI Safety Working Group of ISO/TC22/SC32/WG1423.
The document extends the existing ISO 26262 and ISO 21448 standards, which already address functional safety and the safety of the intended functionality (SOTIF) in the automotive industry. Neither ISO 26262 or ISO 21448 address the specific malfunctions and development lifecycle related to AI/ML algorithms and models used in safety-critical elements used in road vehicles. Therefore, ISO/PAS 8800 proposes a novel AI/ML-based product development lifecycle, which focuses on functional insufficiencies of AI/ML models and classical systematic and random faults or errors of AI-based elements used in road vehicles. This document demonstrates how AI/ML algorithms can be developed in compliance with safety principles defined in both the ISO 26262 and ISO 21448 standards.
What is ISO/PAS 8800?
ISO/PAS 8800 proposes that functional safety-related risks associated with AI system malfunctions can be addressed by tailoring applicable clauses from ISO 26262-4, ISO 26262-6 and ISO 26262-8. Meanwhile, risks arising from functional insufficiencies (performance limitations or specification insufficiencies) of AI systems can be addressed through the extensions of the concepts proposed in ISO 21448 (SOTIF). For this purpose, ISO 8800 proposes a causal model based on a cause-and-effect chain to identify the root causes of the functional insufficiencies.
For example, an incomplete training dataset due to insufficient specification that does not contain a specific type of object (e.g., an animal class) may be encountered during operation in the field which may lead to an animal not being detected or being misclassified by a ML model. In this case, the triggering condition would be the presence of the animal on the road, which may reveal a specification insufficiency that contributes to hazardous behavior on the vehicle level (potential collision with an animal).
ISO 8800 consists of 15 clauses and 8 annexes which cover the following topics:
- Definition of an AI element within the context of road vehicles and AI safety management lifecycle.
- How to develop assurance arguments for AI systems.
- How to adopt a V model for AI systems (for example, deriving of the AI safety requirements, selection of architectural and development measures, verification and validation of AI systems).
- How to deal with data for AI systems.
- How to perform safety analysis for AI systems.
- Which measures should be performed after deployment of AI systems.
- How to demonstrate confidence in the use of AI development frameworks and software tools used for AI model development.
Supporting AI/ML safety, security and quality across the product lifecycle
At UL Solutions Software Intensive Systems, we provide training and consulting services to support your organization in achieving compliance with ISO/PAS 8800:2024. If you use artificial intelligence (AI) algorithms or machine learning trained models (ML) in safety critical functions within cars, we can help you to perform gap analysis to identify the potential areas of non-compliance and develop tailored processes to bridge those gaps.
Our experts participate in various ISO standardization committees, and we have deep expertise in developing AI/ML based products for ADAS (Advanced Driver-Assistance System) and AD (Autonomous Driving) systems for manufactures and suppliers around the world.
With ISO 8800 we can support you in your goals to bring safer and more reliable AI/ML based products to the market.
Read our related content