Overview

This one‑day training course introduces how to bridge traditional functional safety practices with the unique challenges introduced by Artificial Intelligence (AI), whether used within a safety function or as part of an AI‑based development tool. The course is grounded in established functional safety standards such as IEC 61508 and ISO 26262, and incorporates published guidance from ISO/IEC TR 5469 and ISO/PAS 8800, while aligning with the trajectory of emerging international standards, including the upcoming ISO/IEC TS 22440.

The increasing deployment of Artificial Intelligence (AI) in safety-critical systems is fundamentally changing how functional safety must be approached. Traditional safety standards such as IEC 61508 and ISO 26262 were developed around systems whose behavior can be fully specified, traced, and systematically analyzed, enabling predictable identification and mitigation of failures.

AI‑based systems, particularly those using machine learning, introduce behavior that is difficult to predict, explain, or fully specify using traditional deterministic safety assumptions. Their performance depends on data, training processes, and operational conditions, which can vary over time. These characteristics challenge core assumptions of established safety frameworks and require new approaches to hazard analysis, validation, and risk reduction.

In response, industry guidance such as ISO/IEC TR 5469 and ISO/PAS 8800 extends functional safety principles to AI-driven systems and provides a foundation for addressing these challenges in practice. This course equips participants with the frameworks and system-level perspective needed to apply functional safety principles to AI-based systems in real-world development contexts.

The training addresses the specialized approach required for the inherently dynamic nature of AI decision-making. Participants will examine the AI safety life cycle, classification schemes for AI software components, and essential mitigation techniques like diverse redundancy and runtime monitoring. The course also provides a framework for quantifying residual failures and validating non-deterministic software through statistical performance assessments.

Intended Audience

This course is designed for engineers and leaders involved in the development, integration, or oversight of Physical AI systems operating in safety-critical applications. It is particularly relevant for:

• Systems, hardware, and software engineers working on autonomous robotics, humanoid systems, industrial automation, mobile robots (AGVs/AMRs), surgical robotics, unmanned aerial systems, or other embodied AI applications.
• Engineering managers and technical leads responsible for guiding teams through safety-critical development processes for Physical AI products.
• CTOs and technical executives evaluating organizational readiness, risk posture, or strategic direction for responsible robot and Physical AI deployment in safety-critical applications

Objectives

By the end of this course, participants will be able to:

• Understand the fundamental differences between traditionally specified, logic‑driven safety systems and data‑driven AI‑based systems, and the implications for safety engineering
• Interpret how AI technologies are classified within safety-related systems and how this influences safety strategy and required rigor
• Recognize how AI systems are realized across data, training, and inference stages, and where safety risks emerge throughout this lifecycle in alignment with modern safety guidance, including ISO/IEC TR 5469 and ISO/PAS 8800
• Identify key AI-specific risk factors—including uncertainty, environmental complexity, and data-related issues—that challenge traditional safety approaches
• Understand how established functional safety concepts (e.g., HARA, risk reduction, safety functions) from IEC 61508 and ISO 26262 extend to AI-based systems
• Understand how AI faults differ from traditional system failures and how they can manifest across development and operation
• Understand practical architectural and system-level approaches for managing AI-related risks, including supervision, fallback, and monitoring strategies
• Develop an intuition for how safety arguments are constructed for AI-based systems within the context of existing and emerging industry direction, including ISO/IEC TS 22440

Agenda

Below you will find a tentative schedule for this training course.

• The Intersection of AI & Safety: Bridging the gap between traditional functional safety and AI-driven system challenges
• Terminology & Classification: Understanding how AI systems are categorized and how those classifications impact safety strategies and required rigor
• The AI Safety Life Cycle: Extending the V-model to account for data, training, inference, and continuous monitoring, informed by modern and emerging safety guidance, including ISO/IEC TR 5469, ISO/PAS 8800, and the upcoming ISO/IEC TS 22440
• Hazard & Risk Assessment: Integrating AI-related faults into HARA and evaluating risk in systems without fully specified behavior
• AI Fault Analysis: Identifying faults across model development, input handling, and runtime system behavior
• Architectural Mitigations: Designing safety mechanisms, employing redundancy, and managing uncertainty in AI-based systems
• Data Quality & Training: Managing Operational Design Domain (ODD) coverage, addressing bias, and mitigating data and concept drift
• Testing & Validation: Applying statistical performance metrics, robustness testing, and assessing residual risk in non-deterministic systems
• AI-Based Development Tools: Considerations for qualifying AI-assisted tools within safety-critical development environments