Audio Is Never Stored Is Easy to Say. So We Published the Code.

Every ambient AI scribe on the market has a privacy page. A policy describes a promise, but it does not let you inspect how the software handles audio.

We decided to make ours inspectable. We published the code behind palmER's Ambient Scribe as an open-source project on GitHub: github.com/palm-ER/self-hosted-asr.

What we published

Every Ambient Scribe encounter on palmER runs through this code. Anyone can read it, run it, and verify that it never stores audio. The repository includes:

• Speech-to-text that transcribes a complete audio file in one pass
• Speaker identification that labels who said what, from the clinician to the patient or family member in the corner
• Streaming transcription that returns text while the audio is still arriving
• The security configuration and a test suite that exercises the privacy behavior the code claims

If a hospital, a residency program, or a privacy-minded developer wants to run transcription entirely on their own machines, this repository will do it. The code is published under the Apache-2.0 license, and the models carry a CC-BY-4.0 license.

Why we did this

We've examined what AI scribes retain: audio, transcripts, AI drafts, and your edits. Retention depends on contracts, and each type of data may be discoverable. A privacy policy describes a promise; code lets someone inspect the implementation.

In the published stack:

• Audio exists only in memory for the life of the request. Working files exist only while the system is running. When transcription finishes, fails, or the container stops, the audio is discarded. No code path writes it to disk or a database.
• The containers cannot write it elsewhere. Each service runs in a locked-down environment that cannot save new files, and the deployment blocks writes to disk.
• The logs leave patient content out. The system keeps tracking numbers, timing, and error codes, but it does not log audio, transcript text, query parameters, or patient speech.

What this lets you inspect

If you are evaluating an AI scribe for emergency medicine, use the repository as a reference. You can inspect each layer of audio handling instead of relying on a policy page.

For the clinical workflow, the Ambient Scribe user guide explains how to start, pause, and complete an encounter.

Transparency should be the standard

Medical AI creates more data flows than the signed note shows. Patients and clinicians deserve a clear view of how encounter data is handled. Publishing the implementation gives them something concrete to review.

Start your free 30-day trial, read our security page, or review the code directly at the GitHub repository.