
EXPERTISE
Safe & dependable AI.
We assess AI-enabled systems in the role and operating conditions for which they are intended. Perception, localisation and decision-making are considered alongside conventional software, hardware and the mechanisms that limit unsafe behaviour.
Define the safety claim and its limits
We examine operating domains, data and model assumptions, functional insufficiencies and hazardous outputs. The resulting requirements connect AI performance to system consequences and identify where monitoring, operating restrictions or independent protective functions are needed.
Assess statistical validation evidence
Our AI Safety Validation & Dependability Assessment evaluates scenario selection, test sample sizes, dangerous failure-rate estimates and confidence intervals. Rare conditions, correlated failures, distribution shifts and incomplete scenario coverage are examined before drawing conclusions about real-world safety.
Combine AI evidence with system assurance
Statistical results are combined with FMEA, fault trees and deterministic safety mechanisms. We review runtime monitoring, fallback behaviour and the evidence needed after model or data changes, making both supported claims and unresolved limitations visible.
STANDARDS & ENGINEERING REFERENCES
Relevant standards
We agree applicable standards, editions and assessment criteria for your product, operating environment and contractual obligations.
- ISO/IEC 23894
- Guidance on managing AI-related risks and integrating risk management into AI development and use.
- ISO/IEC TR 5469:2024
- Guidance on functional safety and AI systems, including AI-controlled equipment and its safety measures.
- ISO/IEC 24029
- Neural-network robustness assessment: an overview of methods and a methodology for applying formal methods.
- ISO/PAS 8800
- Safety and artificial intelligence in road vehicles: AI safety requirements, evidence and the limits of AI behaviour.
- ISO 21448 · SOTIF
- Safety of the intended functionality in road vehicles: functional insufficiencies, triggering conditions and validation.
ISO/IEC TR 5469 provides guidance. ISO/PAS 8800 and ISO 21448 are automotive references; their methods are adapted where appropriate outside that sector.
These references describe our engineering scope; they are not claims of company or product certification. Browse all standards and their scope.
Bring your engineering question.
Tell us about your system, the decision ahead and the evidence you need. We’ll help define a useful next step.
