Home of the 26262 club

About Automotive SafeAI Day

The Role of AI in Automotive Safety: Developments and Strategic Direction

Join us for an exclusive, invitation-only event exploring the transformative role of AI in enhancing automotive safety. This focused gathering brings together leading functional safety professionals to delve into the latest developments, current practices, and future strategies shaping the integration of AI in the automotive industry.

Through expert-led sessions and strategic discussions, this event provides a unique platform to address the evolving challenges and opportunities in ensuring the safety and reliability of AI systems in automotive applications.

Session Highlights

Session 1: Safety of AI Systems

Examine critical methodologies to demonstrate compliance with EU requirements for freedom from unreasonable risk in AI-reliant designs. This session will also cover advanced safety analysis techniques and explore end-to-end (E2E) approaches to managing AI safety.

Session 2: The use of AI in Development/Lifecycle processes

Discover how DevOps practices are evolving to support safety-critical AI applications in automotive systems. This session will explore necessary process changes for integrating AI into development workflows and the strategic use of AI-driven tools for tasks such as coding and documentation, ensuring alignment with functional safety standards.

Session 3: Strategic Future Planning

Dive into critical areas shaping the future of automotive safety and AI. This session will highlight the importance of robust quality and safety assurance processes to ensure the reliability of advanced electronics in increasingly complex automotive systems.

As a forward-looking discussion, this session offers decision-makers a unique opportunity to reflect on the future of the functional safety profession and make strategic decisions to drive progress in this evolving field.

Lastly, we will take the opportunity to reflect on the activities of the club and get your feedback for future meetings and activities.

Drinks and social agenda will also be part of the day.

How do you show you have met the EU requirement for freedom from unreasonable risk when the design relies on AI?

EU Regulation 2022/1426 establishes type-approval requirements for the automated driving systems (ADS) of fully automated vehicles, ensuring they are free from unreasonable risks to occupants and road users compared to competently driven manual vehicles. Developers face challenges in defining "free from unreasonable risk" in a manner translatable to engineering requirements, validating this definition across all road use scenarios, and demonstrating compliance with these requirements. How can developers ensure that their interpretation of "unreasonable risk" aligns with regulatory expectations and real-world safety outcomes?

Presenter: Roger Rivett

End-2-End AI

Wayve’s innovative approach replaces the modular ‘sense-plan-act’ architecture of the traditional AV1.0 approach with a single neural network trained on diverse data to convert raw sensor inputs into safe driving outputs. This is what we call End-to-End AI (E2E AI). Our technology learns driving skills from raw, unlabeled data using self-supervised learning, eliminating the need to curate expensive and time-consuming labeled datasets. This E2E approach also enables flexible sensor selection, eliminates the reliance on HD maps and can easily adapt to operate on any type of vehicle, from passenger cars to delivery vans.

Presenter: Gareth Price, wayve

Balancing Safety and Innovation: DevOps for ADAS/AD in Delivering Continuous, Safe, and Evolving Products

This presentation focuses on the critical role of DevOps in the development and deployment of Advanced Driver-Assistance Systems (ADAS) and Autonomous Driving (AD) technologies. It will explore how to maintain rigorous safety standards while continuously improving and releasing new features, ensuring both reliability and innovation at every stage of the development lifecycle.

Presenter: Fredrik Sandblom, Zenseact

Challenges for Compiler for AI Models

"In this talk, I’ll explore three critical domains for building confidence in AI-driven systems: functional correctness and completeness (feeding accurate data), robust modeling (using reliable approaches and modeling tools), and transformation accuracy (ensuring compilers do their work). My goal is to spark a discussion on how we, as a community, can address these challenges and possibly establish guidelines for verification and validation. If you’re passionate about functional safety or AI in embedded systems, I’m curious to hear your thoughts. Let’s shape the future of safe AI together!" Stefan-Alexander Schneider

Presenter: Stefan-Alexander Schneider, University of Kempten

Safety of Software-Defined AI-Driven Vehicles

Challenges and key concepts of NVIDIA’s approach to develop safe autonomous systems, covering safety engineering for data-driven systems leveraging design safety, process and tools safety, safety KPIs, and acceleration and scaling.

Presenter: Jonas Nilsson - NVIDIA

Discussion: Is safety and security certification important?

This panel will examine whether safety and security certification, truly adds value to functional safety in the automotive industry. We’ll also explore the impact of AI on safety processes and whether it complements or challenges traditional certification practices.

Moderator: David Ward, Horiba Mira

Discussion: The 26262 club future activities

We dedicate the last session of the day to gathering your feedback and learning about the activities you'd like the Club to focus on in the future. Your input is invaluable, and together, we can shape the 2025 calendar and beyond!