Medical / PHI detection
Enable Medical / PHI named patterns for GenGuard and FileGuard discovery (Agent One 4.0.3572+).
Medical / PHI detection
From Trustity Agent One 4.0.3572 and the matching Cloud Portal, GenGuard and FileGuard can detect medical / PHI-like content using the named pattern pack medical_phi — the same style of pack as credit card, AWS key, and password assignment (not free-form regex).
What it looks for
High-precision signals, including:
- Structured labels such as MRN, patient ID, NPI, ICD-10, HIPAA / protected-health phrasing
- Hebrew operational phrases (for example מספר מטופל, תיק רפואי, אבחנה)
- Clusters of clinical language (discharge summary, prescription, lab result, and similar) — or an ICD-style token next to a clinical neighbor
Benign engineering text without those markers is not treated as a hit.
Where it applies
| Surface | Behavior |
|---|---|
| GenGuard endpoint (Agent One) | Text / file gates on covered apps when Medical / PHI is enabled in Security Policies → GenGuard |
| GenGuard browser | Same named pack when the extension receives policy (via Agent One bridge or portal) |
| FileGuard discovery | Night / Look now passes can propose findings in the medical_phi category for operator review |
For AI tools taking classified files through GenGuard paste / file / dialog gates, see Local AI Runtime (Agent One 4.0.3573+). Hard disk ReadFile deny still waits on the signed driver track.
Enable
- Open Security Policies → GenGuard.
- Under endpoint / browser text patterns, enable Medical / PHI.
- Save. Wait for Agent One policy refresh (or Pull now).
- Validate with a synthetic fixture (fake MRN / clinical wording) — never real patient records in a pilot.
Live Events show the usual Warned or Blocked status depending on the action you configured. FileGuard findings appear under FileGuard discovery for classify / dismiss.