NHS nurses face a critical constraint: you cannot use consumer AI tools (ChatGPT, Claude) to process patient data. It's not compliant with NHS IG Toolkit, Caldecott Principles, or GDPR.
But that doesn't mean AI is off-limits. There are safe, compliant ways to use AI to save time on non-patient-data tasks: studying for CPD, revising clinical protocols, structuring handover notes, and personal learning.
Here's what works—and what absolutely doesn't.
The Hard Rule: No Patient Data in External AI
Before anything else: Do not paste patient information into ChatGPT, Claude, or any consumer AI tool.
This includes:
- Patient names or demographics
- Medical history
- Current medications
- Lab results
- Clinical presentation
Why?
- Caldecott Principles require patient data protection
- GDPR makes you liable for data breaches
- NHS IG Toolkit explicitly forbids external AI processing
- Your trust's data protection officer will not approve it
If your NHS trust has an enterprise AI license (some large trusts have Microsoft Copilot Enterprise with encrypted, on-premise processing), you can use AI in controlled environments. But consumer AI tools are off-limits for patient data.
What You CAN Use AI For (Safely)
Safe use cases:
- CPD and clinical revision (studying guidelines, learning new protocols)
- Protocol understanding (explaining a clinical guideline in simpler terms)
- Handover note structure (help drafting non-identifiable case summaries)
- Personal learning (understanding a disease process, treatment pathway)
- Interview prep (practicing for job interviews, promotion prep)
- Time-off work learning (personal clinical study)
Not safe cases:
- Patient data in any form
- Sensitive NHS documents (specific ward information, incident reports, etc.)
- Screenshots or uploads of NHS systems
- Real case examples with any identifiable information
1. Claude or ChatGPT for CPD and Protocol Revision
What it does: Explain clinical guidelines, NICE guidance, and protocols in clearer language. Help you revise for professional exams.
Practical workflow:
- Open Claude or ChatGPT
- Input: "Explain NICE guidance for sepsis recognition in the ED. What are the key red flags?"
- Claude/ChatGPT provides a clear explanation with clinical examples
- You use this to revise and understand the protocol
What you're NOT doing: You're not putting patient data in. You're asking about clinical knowledge generally.
Time saved: 2-3 hours/week if you're using AI for CPD and protocol revision.
Cost: ChatGPT Free or Pro (£0-15/month), Claude Free or Pro (£0-15/month)
Honest assessment: Both Claude and ChatGPT are knowledgeable about clinical guidelines and NICE protocols. Claude is slightly better at detailed clinical explanations. ChatGPT is slightly better at interactive learning.
For NHS staff: Claude Pro (£15/month) is worth it because of longer context window (you can paste longer clinical guidelines and get comprehensive responses).
Get started: https://claude.ai
2. NHS England AI Lab Resources
What it does: NHS England's AI Lab provides NHS-specific AI tools and guidance. Some NHS trusts have institutional access to NHS-approved AI systems.
Key features:
- NHS-focused AI resources and recommendations
- Trust-level guidance on safe AI use
- Compliant tools approved by the NHS
- Links to ongoing AI in healthcare research
What's available:
- Some larger trusts have enterprise AI licenses (Microsoft Copilot with encrypted processing)
- Smaller trusts may have limited AI access
- All trusts should have NHS AI governance framework
How to access: Check with your trust's IT/digital department. They can tell you what AI tools your trust has approved.
Honest assessment: NHS AI governance is evolving. Your trust likely has an AI policy now (as of 2024-2025). Check your intranet. Most trusts are still cautious about AI, which is appropriate—safety is paramount in healthcare.
Get started: Contact your trust's digital/IT team
3. ClinicalKey AI (Elsevier)
What it does: Evidence-based clinical learning platform with some AI features. Explains medical topics, summarizes guidelines, helps with clinical reasoning.
Key features:
- Evidence-based content (Elsevier journals, textbooks)
- AI explanations of clinical topics
- Clinical decision support (for personal learning, not patient care decisions)
- Mobile app for reading on shift
Cost: Usually provided by NHS trusts (institutional subscription). Some individual subscriptions (£20-40/month).
Honest assessment: ClinicalKey is more trustworthy than ChatGPT for medical information because it's based on indexed, peer-reviewed evidence. However, the AI features are limited compared to Claude or ChatGPT.
If your trust has it, use it. If not, Claude/ChatGPT is better.
Get started: Check if your NHS trust provides access
4. UpToDate AI Features
What it does: UpToDate is a clinical reference tool (most UK trusts provide it). Some UpToDate features now include AI-assisted summaries and explanations.
Key features:
- Evidence-based clinical content
- AI summaries of complex topics
- Clinical algorithms and decision trees
- Mobile access on shift
Cost: Usually provided by NHS trust (institutional subscription)
Honest assessment: UpToDate is the gold standard for clinical reference in the NHS. If your trust provides it, this is your primary learning resource. The AI features are secondary to the core evidence-based content.
Get started: Check your trust's intranet for UpToDate access
5. DailyMed (US) or BNF (UK) for Medication Information
What it does: DailyMed (US) and BNF (British National Formulary, UK) provide medication information.
You could use Claude or ChatGPT to help explain medication information: "What are the key side effects and interactions for Metoprolol?" (no patient data, just general information).
UK context: The BNF is the standard medication reference for UK healthcare. Check BNF first for authoritative UK information. Claude/ChatGPT can help explain complex information but shouldn't replace BNF for clinical decisions.
Honest assessment: Don't use AI for medication decisions in clinical practice. Use it for personal learning and CPD. The BNF is authoritative; AI is supplementary.
Get started: https://www.bnf.nice.org.uk
What About ChatGPT Medical Research?
Some NHS staff ask: "Can I use ChatGPT to help with research?"
If it's patient data: No. Patient data is protected.
If it's methodology discussion or literature review: Yes, carefully. You could use Claude to help structure a literature review or discuss research methodology (non-patient topics). But do not paste real patient data or identifying information.
Better option: Use your trust's research governance team. Many NHS trusts have research ethics and governance support.
Real Workflow Example: CPD Session
Task: You're revising for the CNCC exam (Cardiac Nursing Competency Certification). You need to understand acute coronary syndrome management.
Using AI safely:
- Open Claude: "Explain ACS management according to NICE CG180. What are the key treatment pathways?"
- Claude explains: Pathway for STEMI, NSTEMI, unstable angina, with treatment algorithms
- You read Claude's explanation, cross-reference with BNF and local protocols
- Ask follow-up questions: "What are the bleeding risks with dual antiplatelet therapy?"
- Claude explains the evidence
Time saved: 1-2 hours of revision per topic. If you're studying 20 topics for CNCC: 20-40 hours/year of faster learning.
What you're NOT doing: You're not putting patient data in. You're learning clinical knowledge.
Real Workflow Example: Handover Notes (Non-Patient Context)
Task: You're writing a handover note structure for teaching purposes (not a real patient).
Using AI safely:
- Open Claude: "Give me a handover structure for a teaching handover about a post-MI patient (generic scenario, not a real patient)"
- Claude provides a handover framework
- You use this as a template for real handovers
What you're NOT doing: You're not using real patient information. You're using AI to help structure a teaching example.
Your Professional Responsibility
As an NHS healthcare worker, you have professional responsibility under:
- NMC Code of Professional Conduct (nurses)
- HCPC standards (allied health)
- GMC guidance (doctors)
All of these emphasise patient safety, confidentiality, and data protection. Using consumer AI tools with patient data violates these standards.
Using AI safely for CPD and personal learning? That's fine and encouraged.
Recommended Toolkit for NHS Staff
Free/Low-Cost (Safe):
- Claude Free or ChatGPT Free for CPD and learning (£0)
- BNF Online for medication (usually free via NHS)
- UpToDate (usually provided by trust)
- ClinicalKey (usually provided by trust)
Investment:
- Claude Pro (£15/month) for better explanations and longer context
- ChatGPT Pro (£15/month) as alternative
Cost: £0-30/month (many tools provided by trust). Time saved: 2-4 hours/week on CPD and personal learning.
What to Tell Your Manager
If your manager asks about AI in healthcare:
Safe answer: "I'm using Claude and ChatGPT for CPD and personal clinical learning (explaining protocols, revision). I do not use them for patient data. All patient information stays in NHS systems. This helps me stay current on guidelines and evidence."
Unsafe answer: "I'm using ChatGPT to help with patient care decisions" or "I'm pasting patient information into AI."
If your trust hasn't released AI guidance yet, ask your digital/IT department. They'll provide an AI policy.
Final Recommendation
For CPD and learning: Use Claude Pro (£15/month). It's excellent for explaining clinical guidelines, helping with revision, and supporting professional development.
For medication information: Use BNF (UK standard).
For clinical evidence: Use UpToDate or ClinicalKey (usually provided by trust).
For patient care: Do not use external AI. Use NHS systems and your clinical judgment.
AI tools are helpful for your personal learning and professional development. But patient safety is paramount. Keep patient data in NHS systems. Use AI only for non-identifiable learning.
The time you save on personal CPD can be reinvested in actual patient care—which is what matters most.