Subprocessors

These are every company that receives any of your data, and what each one receives. Our privacy policy describes what we collect and why; this page names who helps us run it.

We do not sell your data.

Our servers and our storage. Everything we keep is kept there, in a private store, encrypted at rest.

Call audio, while listening is on, streamed directly from your browser — it does not pass through our servers. We hand your browser a short-lived key for that, so our own key never reaches your machine.

Your company's compiled documentation; the phrases the transcript produces, so it can judge which of them are questions about your product; the questions it then answers; and the questions and answers you choose to put into a follow-up draft.

Only the text of Emma's answer, so her video can speak it. It never receives your voice, your video, or anything you type — when you speak to Emma, your microphone audio goes only to our speech provider to be turned into text, and the connection to Tavus carries no audio from you at all. Tavus makes no video or audio recording. It does keep its own record of the call — when it happened, how long it lasted, whether it worked — along with the text of Emma's answers, held as the conversation transcript, and its operational logs, under its own policy.

Your IP address and connection details, because your browser connects to Daily directly to receive Emma's video — that is simply how a video call works. It carries Emma's picture and voice to you. It never carries your microphone audio: when you speak to Emma, that audio goes only to our speech provider to be turned into text. Tavus uses Daily to run the call; we never send Daily anything ourselves.

The company name or web address you type when setting up a workspace, and the name of a company you ask for a brief about.

The addresses of your own published web pages, so we can read them.

Your workspace's identifier and the text of the pages it fetched, while it runs the setup work.

The words Emma says aloud in the pre-call huddle.

Your email address, if you join the waitlist on our website. And — if you ask us for a private session with Emma — your name and company, so we can send you the link.

Nothing you type, say, or upload. Cloudflare answers the question “what is the address of emmalabs.ai?” for whoever asks it, which means it sees the fact that a lookup happened — not who you are and not what you asked Emma. It is also named in the mail settings for our domain, so it is part of how email to and from us is routed. It does not carry your call, your questions, or your documents.

The emails Kerushan sends you from his @emmalabs.ai address — so, whatever is in that conversation. This is ordinary person-to-person email, not anything Emma produced on a call. The app’s own automated email goes through Resend, above, not through Brevo.

Zoom transcribes the meeting and sends us the text. We send Zoom only requests to start and stop the transcript stream, and the identifiers and signature that prove the request is from your installed app. We do not send Zoom your questions, Emma's answers, or your documents.

The transcript text for the length of the call, in memory only. That service has no database and no keys to your documents, and writes nothing to disk. Railway keeps its own operational logs of the service; those contain no call content, no participant names and no transcript text.