Artificial Intelligence for Functional Safety
Normative Compliance and Operational Excellence in OHS
A systematic approach to integrating AI into occupational health and safety management systems, in compliance with the most demanding international standards.
Artificial Intelligence — Functional Safety and AI Systems
Framework for using AI in safety functions and ensuring the safety of AI-controlled equipment
E/E/PE Systems (Electrical/Electronic/Programmable) integrating artificial intelligence
Technical Report (TR) - Best practice guide for industry
Multi-sector: industry, automotive, healthcare, construction, critical environments
Definition of safety functions, EUC (Equipment Under Control), classification of AI integration levels
Categorization of specific risks (opacity, non-determinism, dataset shift, adversarial attacks)
Model Creation → Execution → Supervision. Complete traceability of AI artifacts
Robustness, explainability, multi-level verification/validation, mitigation controls, Safety Case
Clear classification of each AI agent's role (observe → advise → authorize → execute) with human supervision adapted to criticality level
RiskAI: drift monitoring, OOD (Out-of-Distribution) detection, sensor validation, real-time inference guardrails
Data→Model→Deployed→Monitored pipeline with safety gates, design reviews, and safety tests at each stage
SafetyGraph: Dataset→Model→Release→SIL claim. Logging of hashes, checksums, SBOM ML, provenance
Third-party verification and validation for critical functions. Mandatory "release to field" gate
HealthAgent: drift surveillance, latency, watchdog. Safety actions (stop, slowdown, LOTO)
AgenticX5's Systematic Methodology
End-to-end traceability: Each stage generates traceable artifacts (SBOM ML, Model Cards, Datasheets) linked in SafetyGraph to ensure compliance and auditability
ComplyAI maintains a Safety Case in Goal Structuring Notation (GSN) directly linked to the Knowledge Graph. Hazard Log, V&V evidence, and requirements traceability.
Non-AI safety monitoring channel to achieve required SIL/PL. Strict separation between AI path and non-AI safety cut-off.
ComplyAI enforces KPIs (robustness, latency, fail-safe). InspectAI verifies field compliance with test reports and scenario coverage.
Defined roles: AI Safety Officer, Model Owner. RACI matrices in SquadraAI. Continuous training and human supervision per EU AI Act Art. 14.
Alignment with ISO 45001, EU AI Act, NIST AI RMF, IEC 61508, Bill 25/GDPR. Automated mapping via SafetyGraph.
Scenario: Vision AI detects worker in risk zone
Compliance: Clause 6 (safety-related AI) + Clause 10 (mitigation)
Implementation: InspectAI with dual-channel, immediate alert + non-AI safety action
Scenario: Prediction of critical equipment failures
Compliance: Clause 7 (traceability) + Clause 9 (V&V)
Implementation: RiskAI with OOD monitoring, statistical validation, stratified sampling
Scenario: PermitAI evaluates risks and recommends controls
Compliance: Clause 5 (integration levels) + Clause 11 (human-in-the-loop)
Implementation: AI in "advise" mode, mandatory human validation, GSN in ComplyAI
Scenario: IoT sensors + AI for air quality, noise, temperature
Compliance: Clause 8 (variable context) + Clause 10 (monitoring)
Implementation: HealthAgent with watchdog, SOSA/SSN sensor fusion, real-time alerts
Proactive detection and AI-based prevention with functional safety guarantees
Facilitated audits, reproducibility of AI decisions, documented evidence for inspections
Seamless integration with existing systems (Procore, Intelex, Cority) via SafetyGraph
ISO/IEC TR 5469:2024, ISO 45001, EU AI Act, IEC 61508, NIST AI RMF, Bill 25/GDPR
AIGovernanceAgent: system registry, compliance assessment, post-market monitoring
Anticipation of future requirements, competitive advantage, stakeholder credibility
Organizations using AgenticX5 benefit from a systematic, auditable approach compliant with the most demanding standards, enabling responsible and secure AI adoption in their OHS operations.
Integrating artificial intelligence into OHS management systems represents a major transformation requiring a rigorous and proven framework.
ISO/IEC TR 5469:2024 provides principles, methods, and processes to safely integrate AI into critical functions.
Normative compliance ensures traceability, explainability, and the ability to demonstrate AI risk control to authorities and stakeholders.
Organizations adopting these standards today position themselves as leaders, anticipate future regulations, and inspire confidence.
An ethical and responsible approach to AI protects workers, preserves the environment, and contributes to the common good.
AgenticX5 embodies this vision: an AI platform for OHS, designed from the outset in compliance with the most demanding international standards.
AgenticX5 - Artificial Intelligence & Functional Safety
For more information on implementing ISO/IEC TR 5469:2024 in your organization:
AgenticX5 & GenAISafety Team
ISO/IEC TR 5469:2024 compliance matrix available
25+ years of expertise in AI & OHS
SquadrAI - OHS & AI Experts
"Responsible AI for Occupational Health and Safety"