Privacy
Last updated 15 October 2026
What leaves your computer, where it is processed, and who is responsible for it. In plain words, and only what is true today.
The short version
You are a diagnostic centre. Your patients are your patients — we process their imaging on your instructions so you can draft a report, and for nothing else.
This page says what we actually do today, in plain words. The full legal policy is drafted and with a lawyer qualified in India; until that review is finished we are not going to publish it as though it were settled. If something here matters to a decision you are making, ask us and we will answer in writing.
What leaves your computer
Before a study is uploaded, the patient identifiers in the DICOM tags — name, patient ID, date of birth, institution, referring physician — are removed in your browser. Sex and age are kept, because they change how a study is read.
Images are checked for identifiers burned into the pixels. Anything that cannot be verified clean is refused rather than uploaded.
The raw files you upload are deleted as soon as the study has been processed. If processing fails they are kept for 24 hours so you can retry without sending them again, then deleted automatically.
Where it is processed
Your account, your reports, the derived images and the patient identifiers are stored in Supabase in the Mumbai region. That part does not leave India.
The AI does. Drafting a report from images, structuring what you dictate, and checking the report for contradictions before you sign are all done by a third-party model that currently runs outside India, so a study’s images and its report text are sent abroad to be processed. We are moving this to a model running inside India; we have no date for it yet, so we are not going to give you one.
Dictation is the exception: speech goes to a recogniser running in India, and the audio and transcript stay there.
What does NOT cross the border with a study is the patient. Names, IDs, accession numbers and referring doctor are held in a separate table and are never part of what is sent — there is a check on the outgoing text that stops the request if one appears in it. What leaves is the image and the clinical text.
You can decline AI assistance on any study and dictate or type the report yourself. Nothing is sent anywhere in that case, and you still get the template, the checks and the signed PDF.
Every model call is recorded with the region it ran in, so this is auditable rather than a promise.
Obstetric ultrasound
Sex determination is a criminal offence in India under the PCPNDT Act. Obstetric ultrasound never reaches an imaging model here. It is blocked at three separate points — the study is routed away from the model entirely, the model is instructed never to state or imply fetal sex, and every generated report is scanned for sex-indicating terms before it is shown. Each block is logged and retained.
What we do with your reports
We use them to run the product for you. We do not sell data, and we do not share it with advertisers.
Separately and optionally, you can consent to your anonymised studies and reports being used to improve the drafting models. That consent is a tick box at sign-up, it is off unless you turn it on, and you can withdraw it at any time from Settings. Withdrawing it also removes your past studies from any future training set — it is not only forward-looking.
Who is responsible for what
For your patients’ data you are the data fiduciary under the Digital Personal Data Protection Act 2023 and we are a processor acting on your instructions. A patient asking about their data should go to you; we have no way to identify them and no relationship with them.
For your own account details — your name, qualification and medical council registration number — we are the fiduciary and you are the data principal.
Asking us something
Write to us and we will answer. Until the reviewed policy is published, that is a person reading an email rather than a form: it is slower and more accurate, which for this material is the right trade.