Privacy

Your data stays with you

A plain statement of what Datathrob collects, what it keeps, and for how long. Last updated October 2026.

← Back to home

Patient data

Datathrob does not collect patient data. The data analysis tools run in your browser, and the files you load are not sent to our servers. Do not type patient names, ID numbers or other identifiers into Ask Nora, into variable names or category labels, or into any text you post or send inside Datathrob.

De-identification, free with every analysis

Before you analyse anything, the Analysis Studio can de-identify your file for you. It flags the columns that identify a person, such as names, medical record numbers and dates of birth, and lets you remove them, replace each one with a study code, or shift dates by a fixed offset so intervals stay correct. The key that links a study code back to a patient is encrypted with a password only you know, and it is saved as a file on your own computer. Datathrob never receives it. The step is free for members, it runs in your browser like everything else, and the audit report records exactly what was removed or replaced.

De-identification reduces risk. It does not make a dataset anonymous on its own, because a combination of ordinary fields can still single out a person. You remain responsible for your institution's rules and your ethics approval.

What we store

On our servers we store your account and membership record: your name, email, username, institution, field, member type, plan, payment reference and the dates of your membership and access code. Your password is held by our sign-in service in a protected form and is never visible to us. Your access code is stored as a one-way hash. Our servers also record how much Nora you use (counts of words processed and an estimated cost), not what you wrote.

Everything you create stays in your own browser, on the device you use: Ask Nora project text (titles, protocol drafts and reference lists), analysis results you save, and the profile notes on your member page. Clearing your browser data removes them, so download what you need.

The Lounge

The Lounge is the one place where we store text you type: your posts, replies, group requests and messages, with your username and the time. Do not post patient data. Each post is checked for email addresses, phone numbers, long numbers and patient-related terms, and a match flags the post for an administrator to review. It does not block the post.

A discussion is archived one month after its last reply. Archived discussions are removed from what members see and kept in an archive that only Datathrob administrators can read. An administrator downloads a copy and keeps it on a Datathrob-controlled server outside Supabase, and may then delete the archived rows from Supabase. Direct messages are stored on our servers, are not part of this archive, and can technically be read by administrators who manage the database. How long messages are kept is still to be confirmed. Lounge notice emails go through Resend, say only how many items wait, and never contain the text of a post or message. You can turn them off on your Lounge card.

Where data is held and who handles it

Our database and sign-in service run on Supabase, with the project hosted in Frankfurt, Germany. The website is served by Cloudflare. Account emails (sign-up confirmation, password reset and access code) and Lounge notice emails are sent through Resend. AI answers come from Anthropic (see below). We add no advertising or tracking scripts to the site. We use your browser's local storage to keep you signed in and to hold your saved work.

Ask Nora and Claude

Ask Nora sends the text you enter, through a Datathrob server function, to Anthropic's Claude service to produce answers. The server function does not store that text. Anthropic handles it under its own terms. The same applies to the results narrative in Data analysis: when you run it, your browser sends summaries of your analysis output (variable names, counts, test results and the cleaning log) to Anthropic through the same server function. It never sends your data rows, and it removes the values of identifier-like columns first. You tick a confirmation box before each run. When Nora searches the literature, the server function sends only the search words and your chosen filters to the databases you select (OpenAlex, PubMed, Europe PMC, Crossref and Semantic Scholar), never your project text. When Nora shows a reference list, your browser checks each DOI with Crossref (api.crossref.org). Only the DOI is sent, never your project text.

Your choices

You can ask us to delete your account and projects at any time. Write to [email protected].

← Back to homePublishing and ethics