Where your data goes
AnswerAside listens to your calls and looks things up in your own files. This page shows what stays on your PC, what is sent out, and who receives it.
At a glance
| Data | Where it goes | Our claim |
|---|---|---|
| Documents | Read from your folders on your PC | Never uploaded |
| Search index & embeddings | Built and stored on your PC | Never leaves your PC |
| Transcripts & summaries | Saved on your PC. Passed to the AI model only to answer a question or write a meeting summary. If you turn it on, a copy is saved to the client’s folder you choose | Never stored by us |
| Audio | Transcribed on your PC | Never leaves your PC |
| Questions + matching snippets | Sent to the AI model (Anthropic, or OpenAI if you choose your own OpenAI key) to write each card | Never stored by us |
| Web searches | Only a generic version of the question | Your files and client names are never sent |
What is stored on your PC
- The documents in the folders you add. AnswerAside reads them on your PC.
- The search index built from them, made by a model that runs on your PC.
- The speech model that turns call audio into text on your PC. It’s downloaded once, before your first call.
- Your meeting transcripts, summaries and answer cards.
One exception, and only if you turn it on for a client in Settings: after each meeting with that client, AnswerAside saves a file with the meeting’s summary and full transcript in a folder inside that client’s folder, so you can open it or point another AI tool at it. If that folder syncs (OneDrive, SharePoint, Dropbox, Google Drive), the file leaves your PC, and anyone the folder is shared with can read it. These files are yours: AnswerAside never uses them in answers or search, and never deletes them.
Apart from that, none of these are stored anywhere else, and your documents and search index are never sent. Parts of your meetings are sent to the AI model only as described below. Each client has its own workspace, and a search during a call only looks in that client’s workspace (plus a General one, if you use it).
What is sent, and when
- While you’re listening: no audio is sent anywhere. The speech model on your PC turns call audio into text as people speak, and the audio is never recorded. The app reports only how many minutes it listened, for the usage counts below.
- When you ask for help: about the last minute and a half of the conversation, plus up to six short passages from your files or past meetings that match the question, go through our servers to the AI model to write the answer card. The client’s name and each passage’s file name, section and date go with them. Your files themselves are not sent, and neither are their locations on your PC.
- When a web search helps: AnswerAside first removes client names, your company name, email addresses, phone numbers and similar details from the question. Only that generic question is sent, in a separate request that can’t see the conversation or your files. If too much has to be removed, no web search is made.
- After a meeting: that meeting’s transcript is sent to the AI model to write a summary, which is saved on your PC. Summaries are on by default; you can turn them off in Settings. The Recap button sends that client’s last one or two meetings in the same way: the summary if there is one, otherwise the transcript, shortened.
Once, before your first call: the app downloads its speech model (about 0.7 GB) from GitHub, at github.com and release-assets.githubusercontent.com. Nothing about your calls or your account is sent with it; like any download, GitHub sees your PC’s internet address. A new version of the app may download a new model the same way.
What we keep on our servers
Your account (name and email), the PCs you’ve signed in on, and usage counts such as minutes transcribed and number of answers. We never store what was said, your questions, your files or the answers. Our logs record only things like which feature was used, how long it took and whether it worked.
If you add your own Anthropic or OpenAI API key, it is kept encrypted on your PC and sent with each answer request. Our servers use it for that request and never save it.
Our providers
- Anthropic writes the answer cards, summaries and recaps, and runs web searches through its search partner, which receives only searches based on the generic question. Anthropic does not use data sent through its API to train its models by default. See Anthropic’s commercial terms.
- OpenAI is used only if you choose your own OpenAI API key in Settings. Then it does everything described above for Anthropic instead: it writes the answer cards, summaries and recaps under your OpenAI account and OpenAI’s terms, and its web search receives only the generic question. We ask OpenAI not to store these requests, and OpenAI does not use data sent through its API to train its models by default. See OpenAI’s data controls for the API.
Our servers run on Vercel, so the text described above passes through Vercel on its way to Anthropic, or to OpenAI if you chose your own OpenAI key. Your account is stored with Neon, and sign-in emails are sent through Resend. GitHub hosts the speech model download. None of them receives anything said on your calls.
Each provider’s own policy covers how long it keeps what it receives.
What AnswerAside never does
- Record or save call audio.
- Join a meeting, speak, or post in the chat. Answers appear only on your own screen.
- Identify people by their voice.
- Mix one client’s files into another client’s answers.