AI is no longer experimental in the automotive world. It’s powering perception, decision-making, and advanced driver functions in production vehicles. But here’s the challenge: AI systems don’t behave like traditional deterministic software, while our safety and process frameworks—like ISO 26262 and Automotive SPICE—were originally built around predictable, specification-driven development.

So how do we bring AI into that world without forcing it into the wrong mold?

With Automotive SPICE v4.0 introducing dedicated Machine Learning Engineering (MLE.1–MLE.4) and ML Data Management (SUP.11) processes, we finally have a structured way to assess ML development within the familiar V-model. At the same time, ISO 26262 continues to define the backbone of functional safety for automotive systems. The real opportunity is not to treat these as separate domains, but to connect them into one coherent assurance story.

This keynote will explore how to practically align AI development with Automotive SPICE and ISO 26262—without adding unnecessary overhead. We’ll look at how ML requirements, data governance, model training, and validation artifacts can be linked directly to safety goals and verification evidence. We’ll discuss why data should be treated as a safety-critical configuration item, and how continuous monitoring and updates can fit into a defensible safety lifecycle.

The goal isn’t just compliance—it’s confidence. By creating a clear connection between AI engineering practices, process capability, and safety assurance, organizations can build intelligent vehicle systems that are innovative, assessable, and trustworthy.