Privacy Policy
What RobotParrot collects, why, and exactly how long we keep it.
Last updated: June 20, 2026
RobotParrot turns a resident's dictated case into a personalized audio lecture. To do that we have to handle your recording, your account details, and the lecture we generate for you. This policy explains every category of information we capture, how it is used, who processes it on our behalf, and the specific schedule on which it is deleted.
The short version. Sign out your case as "the patient" — never by name, MRN, or date of birth. We strip identifiers from every transcript before it is stored. Your original recording is deleted after one year; your lecture audio and downloadable transcript are deleted after 30 days unless you pin them. RobotParrot is a learning tool, not a medical record — always follow your own institution's policies on protected health information.
1. Who we are
RobotParrot is a product of Bramble Tree Property, LLC, based in Orlando, Florida. You can reach us about anything in this policy at hello@robotparrot.ai. RobotParrot is currently in private beta.
2. What we collect
Account information
When you sign in, our authentication provider (Clerk) gives us your email address and a unique account identifier; we also store any display name you provide, the times you sign in, your remaining lecture tokens (credits), your email preferences, and — if you subscribe to your private lecture podcast feed — an opaque token that lets your podcast app fetch your episodes.
Your case recording and the lecture we make from it
Each time you record a case, we capture and process:
- The audio recording itself — the file you dictate, plus its format and length.
- A transcript of your recording — produced by speech-to-text, then immediately de-identified (see Section 4). Only the de-identified transcript is stored.
- The structured case we parse from the transcript (presentation, workup, decisions, diagnosis, and the points you said you were unsure about), the cited research assembled for it, and the final lecture script and the lecture audio (MP3) we generate.
- Operational metadata — pipeline status, timestamps, and the per-lecture processing cost, used to run the service and support your account.
Feedback
If you use the in-app feedback button or rate a lecture, we keep your rating and any voice note (and its transcript) so we can improve the product during the beta.
What we do not collect
We do not ask for patient identifiers, and the guided prompts never request them. We do not sell your data, we do not use it for advertising, and we do not use it to train third-party AI models (see Section 6).
3. How long we keep it — retention schedule
Different items have different lifetimes. This is the schedule we operate to:
| What | How long we keep it | Then what |
|---|---|---|
| Your original case recording (the raw audio) | 365 days from recording | Permanently deleted from storage. Kept this long only so we can review lecture quality against the source; never longer. |
| Your generated lecture audio (the MP3) | 30 days from creation | Permanently deleted — unless you pin it (see Section 5). |
| Your downloadable transcript file | 30 days from creation | Permanently deleted — unless the lecture is pinned. |
| The de-identified case text and lecture script (your learning record) | Kept as part of your account | Retained as your de-identified case library. It contains no patient identifiers. Deleted when you delete your account or on request. |
| Account information (email, name, sign-in history, tokens) | For the life of your account | Deleted, or anonymized, when you close your account or ask us to. |
| Feedback (ratings, voice notes, transcripts) | Duration of the beta | Retained to improve the product, then deleted or anonymized. |
| Shared clinical research (built per diagnosis, contains no patient information) | Rolling ~30-day refresh | Regenerated periodically. This is topic-level evidence about a diagnosis — it is never linked to you or any patient. |
Once an item passes its window, access to it is cut off immediately and it is removed from storage on our automated deletion schedule.
4. How we protect patient privacy
RobotParrot is designed so you never need to share a patient identifier — and it removes any that slip through before anything is stored:
- Two-pass de-identification runs before any transcript is saved or sent to the lecture models. A pattern pass removes formatted identifiers (medical record numbers, dates of birth, phone numbers, addresses, salutation-plus-name). An AI pass then replaces any residual names or places with role descriptions ("a 54-year-old," "an outside hospital").
- The identifiable text exists only briefly in memory during this scrub. What is written to storage is already de-identified.
- Your raw recording can still contain spoken identifiers, which is why it is handled purely through the short retention window in Section 3 rather than kept around.
This de-identification is aggressive by design and will occasionally over-redact a clinical detail. RobotParrot is a learning tool, not a medical record or a covered HIPAA service — always follow your own institution's policies on protected health information.
5. Your controls
- Pin a lecture — keep up to five lectures (audio and transcript) indefinitely, past the standard 30-day window. Unpin at any time to return them to the normal schedule.
- Delete — you can ask us to delete a specific case, or your entire account and everything tied to it, at any time by emailing hello@robotparrot.ai.
- Email preferences — you can opt out of non-essential email.
- Access — you can request a copy of the information we hold about you.
6. Who processes your information
We use a small set of vetted service providers to run RobotParrot. They process data only to deliver the service to you, under their own security and confidentiality commitments:
- Render — application hosting and database (United States).
- Cloudflare R2 — encrypted storage for audio files and transcripts.
- Clerk — sign-in and account management.
- OpenAI — speech-to-text (transcription) and text-to-speech (the lecture voice).
- Anthropic (Claude) — the AI agents that research, draft, and check your lecture.
Content sent to our AI and transcription providers is sent for the sole purpose of producing your lecture and is not used to train their models. We do not sell or rent your information to anyone.
7. Security
Access to the app requires sign-in. Audio and transcripts are stored in private, access-controlled storage and served only through short-lived, signed links. Internal admin tools that pair a recording with its lecture (for quality review) are restricted to RobotParrot operators. No system is perfectly secure, but we work to keep the amount of sensitive data we hold — and how long we hold it — as small as possible.
8. Children
RobotParrot is intended for licensed clinicians and medical trainees and is not directed to anyone under 18.
9. Changes to this policy
We may update this policy as the product evolves. When we make a material change we will update the date at the top and, where appropriate, notify you.
10. Contact
Questions, access requests, or deletion requests: hello@robotparrot.ai.