Redact patient data before AI — now on your iPhone.

Clinicians and medical writers already paste notes into ChatGPT and Claude — the tools are too useful not to. But that text is dense with identifiers, and once it leaves the device you've lost control of it. Redacta closes that gap: it finds the identifiers and replaces them with neutral tokens like [NHS_NUMBER_1] before the text reaches any AI, so the meaning stays intact and the patient doesn't.

It's a native iPhone app. Detection runs entirely on the device using the same engine that powers the Redacta skill, MCP server and libraries — there is no network connection, no account, and no analytics. Its App Store privacy label is, truthfully, Data Not Collected.

The token map that reverses a redaction is held in memory for the session only and is never written to disk — so the round trip works without anything being stored.

Three modes — Clinical, General PII, and HIPAA Safe Harbor — with UK-grade detection: NHS numbers validated by the Modulus-11 checksum, National Insurance numbers, dates of birth, postcodes, hospital/MRN numbers, emails and phone numbers. Clinician names are preserved by design; only the patient's data is removed.

Privacy isn't a setting in Redacta — it's the architecture. There is no networking code in the app at all, so there's nothing to switch off and nothing to trust: turn on Airplane Mode and it works exactly the same. No account, no sign-in, no analytics, no tracking.

The text you redact, the photos you scan, and the token map that reverses a redaction are all processed locally and never transmitted or stored. That's what lets the App Store privacy label read, truthfully, Data Not Collected — and why Redacta is a genuine fit for work bound by confidentiality and data-protection rules.