Practice position
AI assists. People decide.
Use approved systems, minimise information, check important output and keep accountable professional judgement human-led.
Keep consequential decisions human-led
Do not ask general-purpose AI to make a safeguarding decision, diagnosis, prescribing choice, individual risk assessment, treatment-eligibility decision or decision to restrict a service. These require accountable roles, evidence, context and established processes.
Do not use identifiable records casually
Client records, health information, criminal-justice information and safeguarding details should not enter consumer or unapproved AI systems. A local approved use requires formal information governance, supplier assurance and defined purpose.
Do not confuse support with authority
AI can help prepare a blank checklist or summarise public guidance. It cannot replace supervision, emergency routes, professional consultation or reading the authoritative source.
Practical checklist
- Does this affect care, safety, liberty or access?
- Is it a clinical or safeguarding judgement?
- Does it require identifiable information?
- Can a qualified person independently check it?
- Is there a safer established process?
Authoritative starting sources
WHO guidance on generative AI for health ↗NHS England AI information governance ↗Written by Tony D’Agostino / TD Consultancy. Reviewed . This page provides general professional education, not legal or clinical advice.