Where do you start
with ISO/PAS 8800?

Average model performance cannot answer every question about deploying AI in a vehicle. Engineers also need to consider its operating context, the vehicle-level effects of incorrect outputs, and the system’s response.

The specification’s starting point

ISO/PAS 8800:2024 is titled “Road vehicles — Safety and artificial intelligence”. It addresses safety risks arising from AI output insufficiencies, systematic errors, and random hardware errors in relevant vehicle electrical/electronic systems.

Its scope focuses primarily on machine learning. It does not provide specific guidance for software development tools that use AI methods.

Four questions to organize your thinking

1. What is the AI function?

   Describe the inputs, outputs, intended purpose, and system boundary.

2. When could it behave inadequately?

   Consider operating conditions, data limitations, and the effects of inadequate outputs.

3. How does the system respond?

   Explain monitoring, reference paths, alternative behavior, and their limitations.

4. What evidence supports the argument?

   Examine how requirements, design decisions, and verification results connect.

These questions are an educational starting point proposed by the forum. They do not replace the complete specification or a project-specific conformance assessment.

Begin with a concrete architecture

Our autonomous-vehicle and battery-AI research notes provide concrete designs for exploring these questions. Review the scope of demonstrated findings and architectural proposals as you work through them.