The data flow

What happens when you generate a brief

1. You fill in the form (or upload a document), then click Generate.
2. Your input is sent onceto Anthropic’s API so Claude Opus 4.7 can read it and produce the brief.
3. The brief streams back to your browser and renders.
4. That’s it. We do not save the brief, the input, or the document anywhere on our servers. There is no database. There are no accounts. When you close the tab, the brief is gone unless you printed or saved it.

The fine print, defined

Specifically:

• We do not retain any input you provide. Not in a database, not in a log, not in a file.
• We do not require an account. No sign-up, no email collection, no profile.
• We do not use your data for training. Anthropic explicitly does not train its models on customer API inputs (per their commercial terms).
• We do not share your data with anyone other than Anthropic, who processes it only to generate the brief and returns it.
• We do not run analyticson what people enter or generate. We don’t know what you typed, we don’t know what brief you got back.
• We do not place tracking cookies. The site is deployed on Vercel, which collects standard server logs (IP, user agent, timestamps) for security and abuse prevention. None of that data is tied to your form input.

When you upload a document

Documents (PDFs, images) and how vision works

When you drop a PDF or image into the upload zone, the file is base64-encoded in your browser, sent to Anthropic’s API as part of the request, parsed by Opus 4.7’s vision capability, and then the response streams back. The file is not stored by us after the request completes.

Anthropic’s API processes the document to generate the response and does not retain it for training. Their commercial terms apply: anthropic.com/legal/commercial-terms.

What we are not

We are not a regulated medical device

Second Opinion is a research-helper tool. It does not diagnose, treat, or provide medical advice. It is intended to support a shared-decision conversation with your licensed healthcare provider, not replace it. Every brief carries this disclaimer.

We are not a Health Information Custodian under PHIPA (Ontario) or a Covered Entity under HIPAA (United States), because we are not collecting, storing, or transmitting health information on behalf of a clinical organization. You as the patient voluntarily provide the input.

Your control

You are in charge of what you share

You can use Second Opinion without entering any identifying information. The model only needs clinical context (symptoms, imaging text, treatment history) to generate a useful brief. Names, dates of birth, MRN numbers, and doctor names are not required and are not used in the reasoning.

If you upload a document that contains identifying information, the system prompt instructs Opus 4.7 to extract clinical content only and to strip identifiers from its reasoning.

Questions

Want to talk about this?

Email dev@jamesdare.com. The project is built and maintained by James Olusoga in Toronto. The source code is open and reviewable on GitHub.