O3
GUARDRAILS FOR AI IN OCCUPATIONAL HEALTH USE: GUIDANCE FRAMEWORK AND RISK CHECKLIST
- G. O’Neill, PAM, Burton Upon Trent, UK
Background
AI use in occupational health (OH) is expanding, yet services lack practical, sector‑specific guardrails to ensure safe, lawful, and effective deployment across clinical and non‑clinical workflows.
Aims
To co‑produce a concise guidance framework and an operational risk‑assessment/software as a medical device (SaMD) checklist to support UK OH professionals in responsible AI adoption aligned with legal, ethical, and regulatory expectations.
Methods
We created requirements from data protection, equality, product safety law, and health sector standards into an AI Use Guidance for UK OH Professionals, including transparency, fairness, privacy, human oversight, accountability, risk management, training, incident reporting, and sustainability. Included is a practical checklist with scoring and an SaMD triage aligned to validated guidance.
Results
The guidance provides structured principles, procedures, and resources for OH teams, including documentation expectations, appeals processes, and links to authoritative templates. The SaMD component offers a decision aid spanning intended purpose, classification, clinical safety (DCB0129/0160), data protection, security, bias, accuracy, training, and governance. The risk tool introduces severity and likelihood scoring, risk logs, and review cadence to support ongoing monitoring and assurance. Governance features include a risk register, regular reviews, stakeholder involvement, and practical sustainability prompts (necessity, energy use, and right‑sized models).
Implications for policy and practice
Co‑produced guidance and tools translate high‑level AI governance into actionable OH practice, enabling safer use and oversight while building workforce capability. Planned next steps are piloting, evaluation, and periodic updates as regulation and best practice evolve.
References:
- Information Commissioner’s Office. AI and Data Protection Risk Toolkit (v1.1), 2025.
- HSE. Regulatory approach to Artificial Intelligence (AI), n.d.
- MHRA. Software and AI as a Medical Device: Guidance, 2025.
- NICE. Evidence Standards Framework for Digital Health Technologies, n.d.
- Regulation (EU) 2024/1689: EU Artificial Intelligence Act, 2024.
