Clinical safety
AI triage should be assessed against urgent, ambiguous, complex and routine scenarios, with documented hazards and controls.
← Sitora Research
AI, Safety & Governance · July 2026
AI as the NHS Front Door
Safety, accountability and public trust in AI triage.
This Sitora policy report examines the safeguards needed when AI influences how patients are directed through NHS services, including clinical safety, independent audit, human escalation, transparency, accessibility and ongoing monitoring.
Sitora position
AI can help the NHS, but AI triage should be clinically safe, independently audited, transparent, accessible and accountable before it becomes a major front door to public healthcare.
Independent assurance
Large-scale deployment should not depend solely on supplier assurances. Independent review should test safety, bias, accessibility and failure modes.
Human escalation
Patients should have a clear route to human review where symptoms are serious, worsening, complex, uncertain or disputed.
Defined accountability
Responsibility between NHS organisations, clinical safety owners, commissioners and technology suppliers should be clear before deployment.
Patient transparency
Patients should know when AI is being used, what role it plays, what it cannot do and how to challenge or escalate an outcome.
Ongoing monitoring
Deployment should include monitoring for incidents, near misses, complaints, demographic disparities, model changes and performance drift.
The Sitora NHS AI Trust Test
The report proposes six public-facing assurance questions: clinical safety validation, independent AI audit, human escalation, defined accountability, patient transparency and continuous post-deployment monitoring.
The Trust Test is not intended to replace NHS clinical safety standards, NICE evidence frameworks, MHRA regulation, ICO requirements or CQC oversight. It is intended to translate those responsibilities into a clearer assurance model for NHS boards, policymakers and patients.
Read the complete report
The PDF includes the evidence base, regulatory landscape, risk analysis, policy recommendations, implementation roadmap, risk register and assurance checklist.
Download full PDF