Your data, handled with care.

• Voice input. When you dictate, audio is streamed to a transcription provider and processed in real time. Audio is not stored after transcription.
• Transcripts. In Private Session (default), transcripts are processed in memory and are not stored in our database.
• Account info. We store basic account details (such as email) for authentication.
• Content you choose to save. Some features exist to keep something for you, and those things are stored on your account: pearls you save, cases you flag for follow-up, your learning progress and streaks, your answers to reasoning challenges, and your settings. Each one is the result of a deliberate action — nothing from a case is kept unless you save it.
• Teaching cases you author. If you create a teaching case, the case text and the de-identified images you upload are stored on your account. Images are held with a third-party cloud storage provider and are served only to people authorized to view that case.
• Reasoning capture, when you turn it on. Tracking your reasoning over time is off by default. If you enable it, we store your own differential before and after each case, plus a short case description you write, so the progress view can show how your reasoning is developing. It stays visible to you alone unless you separately consent to a research study.
• Usage metadata. We store non-identifiable usage data such as model usage, latency, and token counts to maintain and improve the service.

• We do not store dictated audio.
• We do not store transcripts in Private Session.
• We do not log transcript text (it is redacted before logging).
• We do not train AI models on your inputs — every model vendor is configured under no-training agreements.
• We don't sell your data. Ever.
• We don't run advertising trackers or share your activity with ad networks.

Voice and transcripts are used solely to generate your differential and reasoning output during the active session. Aggregate usage signals are used to monitor uptime, debug failures, and improve model selection. That's it.

We minimize data collection by default. In Private Session, no transcript data is persisted. Operational metadata is restricted to a small set of authorized operators under least-privilege access controls.

You can request export or deletion of your account and associated records anytime by emailing support@radreason.com. The active in-app session itself is local — leaving the page, logging out, or starting a new case clears the on-screen conversation from your device.

Retention. Saved content — pearls, teaching cases and their images, progress, and reasoning capture — is kept until you delete it or ask us to. We don't expire it on a timer, and we don't delete it without being asked. Individual items can be deleted in the app; for a whole account, email the address above.

RadReason is designed for de-identified clinical reasoning. In Private Session, no transcript or case text is stored. However, users should avoid entering direct patient identifiers (names, MRNs, dates of birth, accession numbers).

RadReason is not a HIPAA-covered service unless a Business Associate Agreement (BAA) is in place.