Everything below is the short version of the privacy policy, written for the questions people actually ask: where does my study material go, what can a connected assistant reach, and what happens when I delete something.
A study sends the material you are testing, plus the audience context for it, to the models that generate panel responses — Amazon Bedrock, OpenAI, Google Vertex AI or Anthropic, depending on the models configured for the study. Interviewing a respondent sends more: your question, the transcript so far, and the response you are asking about. Your studies, results and uploads are stored on AWS in the United States. Billing goes through Stripe, which holds the card details; Mirror Panel never sees them.
Connecting Claude or ChatGPT gives that assistant the same access you have: your studies, results, brand settings and assets, for the brands you can already see. Disconnecting it in the assistant stops it making further requests, but a credential it already holds keeps working until it expires — up to 12 hours for an access token, and up to 30 days if it kept a refresh token. Tell us if you think a credential has been misused.
We record sessions with Microsoft Clarity — pages, interactions, replays and heatmaps — so a reported problem can be traced to the session that produced it. A signed-in session is labelled with your email address and display name. There is no in-product opt-out yet; a tracker-blocking extension stops it, or ask support to exclude you.
The full privacy policy is at /privacy and the terms at /terms. This article summarises them; where the two differ, the policy is what governs.