Sign-in is Google only, restricted to verified accounts on the modulate.ai Workspace domain. There are no passwords to leak. Sessions expire after two weeks; sign out ends them immediately.
You see your own interview. Project owners (the people running the research) see every interview in their project. A project can additionally open results to all its participants; that setting is visible on the project page before you start.
Everything is stored in a single database on an encrypted volume at Fly.io, in one region, reachable only by this application. Uploaded files are stored in the same database and served only to you and the project owners. There is no analytics or third-party tracking on these pages, and transcripts are never written to server logs.
The project owners write a detailed interview guide: what to ask, in what order, how to follow up and what not to do. The interviewer follows that guide.
The interviewer is an AI model (Claude by Anthropic) accessed through Anthropic's API. Your messages and uploads are sent to Anthropic to generate the interviewer's replies. Anthropic does not train on API data; requests are retained for up to 30 days for abuse monitoring and then deleted. No other AI provider is involved, and the model has no web access.
Project owners can delete a single interview or the entire project, which removes transcripts, answers and files immediately. Ask the project owner if you want your interview removed.