Documentation

Last reviewed: May 2026

Patient Data Protection in AI — Demonstration

This page provides a short recorded demonstration of the patient-identifiable information (PII) control described in our Patient Data Protection in AI policy. It shows the control working in the live product: a prompt containing patient-identifiable information is detected and blocked before it can reach the AI model.

Synthetic data only

The recording uses synthetic test data only. No real patient-identifiable information is shown at any point.

Download the recording (MP4)

What the demonstration shows

In the recording, a user enters a prompt that includes patient-identifiable information into Widgi AI. Before the prompt is sent, the detector identifies the patient data, classifies it as high risk, and blocks the prompt — it is never passed to the AI model. The user is prompted to remove the personal details and use a generic, non-identifying description instead.

What this evidences

The demonstration confirms the key control claim: high-risk, patient-identifiable prompts are blocked before they reach the AI, and the user is guided to resubmit without personal data. Full technical detail of how the control works — deterministic rules-based detection, NHS check-digit validation, browser- and server-side enforcement, and category-only logging — is set out in the Patient Data Protection in AI policy.