Deliberate source selection
Choose a local call/recording, current Symptom Checker session or ended Wellness session rather than importing everything automatically.
Clinical demonstrator
Local product demonstratorMeridian Risk Guardian
Bring deliberately selected conversation and synthetic-record context together for a human-reviewed local signal assessment.
Selected evidence. Transparent review signals.
Risk Guardian is a local review prototype for inspecting selected evidence. A reviewer chooses each source, checks the imported context and runs configured rules or separately permitted model observations. It is distinct from Guardian’s operational service-supervision workspace.
What the AI does. What people decide.
The actual assistance
Configured phrase rules and weights create transparent review signals. Optional model-assisted semantic observations require separate authorisation and consent for the selected wording. A result is not a validated clinical-risk score or urgency decision. No matches must not be interpreted as proof that somebody is safe.
Human ownership
No diagnosis, clinical risk classification, urgency assignment or care decision is established. Keep unvalidated rule sets disabled for live use.
Features in the current workspace.
Explicit patient context
Import an actively selected synthetic Patient App record or consented handoff only after verifying that it is the intended record.
Editable evidence inputs
Inspect symptoms/caller wording, record notes, voice analysis, semantic observations and other context before review.
Authorisation confirmation
Confirm authority to review the selected local sources; model analysis has a separate opt-in.
Transparent rule calculation
Run configured phrase rules and weights and inspect the resulting review signals.
Optional semantic analysis
Invoke the configured Meridian model only when enabled and permitted for the selected wording.
Human review record
Record a reviewer note and local review status, while keeping clear that this prototype review is not persisted by Risk Guardian.
Rule configuration
Define local categories and exact phrases, add rules and save configuration; rules are not validated clinical thresholds.
Source clearing
Clear evidence and keep the fact that raw audio is not imported visible.
From first input to a reviewed next step.
- 01
Select a source
Choose each local item deliberately.
- 02
Check identity and authority
Verify the selected synthetic record and permitted purpose.
- 03
Review inputs
Edit or remove irrelevant evidence.
- 04
Run the chosen analysis
Use explicit rules or permitted optional model observations.
- 05
Interpret responsibly
A person decides what the signal means and what follow-up is appropriate.
How the product works.
| Area | Mechanism and scope |
|---|---|
| No name matching | Patient records are not automatically linked to a call by a similar name; selection and verification remain explicit. |
| Local signals | Exact phrases, categories and weights produce advisory results rather than clinical thresholds. |
| Model opt-in | Configured model analysis is separately consented and uses selected wording. |
| Persistence boundary | This Risk Guardian prototype does not save the imported evidence or local review record; case-management review summaries have a distinct confirmed save path. |
Product evidence and source notes
Content checked against the v25.2.0 source and visible local workspace on 30 October 2026. This is a product description and interface inspection, not an end-to-end deployment or clinical validation.
static/risk-guardian.js; current workspace inspection; case-management review boundary
Review assurance scopeQuestions about Risk Guardian.
Is Risk Guardian the same as Meridian Guardian?
No. Risk Guardian is an explicitly selected evidence-review prototype. Guardian is the operational supervision product.
Does importing a patient record automatically match it to the call?
No. A person selects and verifies the intended synthetic record; the tool does not establish patient identity.