A safety case is a structured, evidence-backed argument that a system is acceptably safe to operate in a defined context and operational environment. It comprises explicit safety claims, the reasoning that connects them, and the supporting evidence drawn from hazard analysis, design measures, verification, and operational data. Commonly documented using notations such as the Goal Structuring Notation or Claims-Arguments-Evidence, the safety case is reviewed by stakeholders and regulators and maintained throughout the system lifecycle. It is the central assurance artefact in safety-critical and autonomous-systems engineering.
Overview
- The safety case answers a deceptively simple question: why should we believe this system is safe enough? It makes the reasoning explicit rather than implicit, so reviewers can challenge claims, inspect evidence, and identify gaps.
- Why it matters: in Autonomous Systems, medical robotics, automotive, rail, and aviation, regulators require a defensible, documented assurance argument before deployment. The safety case provides that argument and is maintained as the system evolves.
- How it works:
- Safety goals are derived from hazards and the associated risk-reduction targets.
- Each goal is decomposed into a strategy and supporting sub-goals.
- Evidence (analyses, test results, field data, process records) is attached to the lowest-level claims.
- The argument is reviewed, and any defeaters or counter-evidence are addressed before approval.
Key aspects
- Claims — the assertions about safety properties the system must satisfy.
- Argument — the explicit chain of reasoning connecting claims to evidence, often expressed in Goal Structuring Notation (GSN) or Claims-Arguments-Evidence (CAE).
- Evidence — verification and validation results, Fault Tree Analysis, formal analyses, simulation, and operational monitoring data.
- Confidence and counter-evidence — assurance deficits and defeaters must be surfaced and resolved, not hidden.
- Lifecycle maintenance — the safety case is a living document updated as the design, environment, or operational data change.
Applications
- Automotive systems certified under ISO 26262, including driver-assistance and autonomous driving functions.
- Personal-care and service robots assured against ISO 13482.
- Rail, aviation, nuclear, and defence systems where regulators mandate a documented assurance argument.
- Medical devices and surgical robots requiring Regulatory Approval before market access.
- Increasingly, assurance of machine-learning components within safety-critical autonomous platforms.