Practice position
AI assists. People decide.
Use approved systems, minimise information, check important output and keep accountable professional judgement human-led.
Do not delegate the judgement
Safeguarding requires context, professional curiosity, lawful information sharing, relationships, escalation and accountability. A general-purpose AI system cannot take responsibility for whether a concern meets a threshold or what immediate action is needed.
Do not enter identifiable safeguarding information into an unapproved tool or ask a chatbot whether a person is safe.
Lower-risk support around the work
AI may help create a blank briefing structure, fictional training scenario, public-information draft or generic questions for a policy review. Those outputs still need safeguarding and information-governance review.
Follow established routes
If there is a live concern, use organisational safeguarding, emergency and supervision procedures. AI should never delay contact with the accountable person or service.
Practical checklist
- Use established safeguarding routes
- Do not enter identifiable concerns into general AI
- Keep thresholds and actions human-led
- Verify any educational content
- Record and escalate errors through existing processes
Authoritative starting sources
NHS England AI information governance ↗WHO generative-AI governance for health ↗Written by Tony D’Agostino / TD Consultancy. Reviewed . This page provides general professional education, not legal or clinical advice.