Evidence
Every claim needs a source, a grade, and a versioned review trail.
Evaluate input
No evaluation yet.
Privacy
Consent, export, correction, and deletion flows are required.
Safety
Allergen, interaction, pregnancy, paediatric, and contraindication rules must gate outputs.
What must be blocked or downgraded
- Diagnostic claims
- Prescribing instructions
- Emergency triage beyond referral
- Unverified health claims without evidence
- High-risk advice for children, pregnancy, or medication interactions without review
Governance stack
Use the imported extensions as the policy baseline: threat model, privacy/DPIA, clinical evidence governance, testing/evaluation, CI/CD supply chain, source governance, and incident runbooks.