Earshot

Last updated 19 August 2026

Privacy Policy

Earshot reads your mailbox and records your meetings. There is no version of that which is not sensitive, so this page says exactly what is collected, where it goes, who else can see it, and how to take it back or destroy it. It is written to be read rather than to be survived.

The short version. Audio never leaves your machine. Mail is read and never written: Earshot's access to Gmail is read-only and it cannot send anything. Nothing you give Earshot is sold, shared with other users, or used to train anybody's model. You can export everything and you can delete everything, from inside the app, without asking us.

Who we are

Earshot is operated by the maintainers of earshotcrm.com, who are the data controller for the information described here. Write to hello@earshotcrm.com about anything on this page, including a request made under GDPR, the CCPA, or any equivalent law.

What Earshot collects

WhatWhere it comes fromWhere it is kept
Your account — email address, and a Google account identifier if you signed in with Google You, when you sign in Firebase Authentication
Email metadata and content — sender, recipients, subject, date, thread, and the body text of messages Earshot judges to be correspondence Gmail, only after you connect a mailbox and only as far back as the window you choose Firestore, in your account
Calendar events — title, time, location, and attendees of upcoming meetings Google Calendar, read-only, read live Nowhere. Read at the moment a screen is drawn and never stored
Recordings — the audio of meetings you record Your microphone and your Mac's system audio, only while you are recording A folder on your Mac (~/Music/Earshot). Never uploaded
Transcripts and write-ups — the words of a call, its summary, and who it was with Produced from your recordings Your Mac, and Firestore in your account
Contact profiles — what Earshot has worked out about the people you deal with, from their mail and your calls with them Produced from the above Firestore, in your account
Usage counters — how many minutes of audio and how much model time your account has used this month Produced by using Earshot Firestore, in your account
How long Earshot's own jobs took — for each time it reads your mailboxes, how many seconds each stage took, how many messages there were, which provider they came from, and how many requests and retries it needed. Counts and durations, and nothing about the mail itself Produced by using Earshot Firestore, filed under your account id

Earshot has no advertising identifiers, no session recording, and no third-party trackers or analytics services on this website or in the app. It does not collect your location, your contacts list, your browsing, or anything about what else is running on your computer.

It does time itself. Reading a mailbox takes minutes, and the only way to make it take fewer is to know where the minutes go — so when Earshot finishes reading your mail it records how long each stage took, how many messages there were, and how many requests it made, and sends that one summary to its own server. No address, no subject, no message, no name, and no domain is in it: it is a row of numbers plus the word "google" or "microsoft". Nobody else receives it, and it is never used for anything but making the app faster and working out what it costs to run.

What Earshot does with it

One thing, and only this thing: it builds and keeps the picture of your working relationships that the product is for — who you know, where each thing stands, and what you owe them. Every use of your data below is in service of that.

Your content is never used to train a model — not ours, and not by the providers below, whose enterprise terms forbid it. It is never sold, never rented, never shared with other Earshot users, and never combined with any other account's data.

Audio, specifically

Recordings are files on your machine. They are not uploaded, not backed up by us, and not stored anywhere but the folder you can open in Finder. When a recording is deleted in Earshot it is moved to your Trash, and what happens to it after that is between you and your Trash.

Transcription is the one exception, and it is worth being exact about. Turning speech into words requires a model to hear it. If transcription is set to the cloud backend, chunks of speech — silence removed, downsampled to 16 kHz mono — are sent to OpenAI, transcribed, and returned. Nothing is written down on the way: the chunk exists for the length of one HTTPS request and is not stored by Earshot's API. If you set transcription to run on your own machine instead, no part of any recording leaves it, ever.

Google user data, and the Limited Use requirements

Earshot asks for three Google scopes. All three are read-only. Earshot reads from your Google account and writes nothing back to it: it cannot send, reply to, compose, alter, label, archive or delete anything, and it holds no permission that would let it. Google lets you untick any of them at the consent screen, and Earshot degrades rather than breaks when you do.

Earshot's use and transfer of information received from Google APIs to any other app will adhere to the Google API Services User Data Policy, including the Limited Use requirements. Specifically: Google user data is used only to provide and improve the user-facing features described on this page; it is not transferred to others except as necessary to provide those features, to comply with applicable law, or as part of a merger or acquisition; it is not used for advertising; and no human reads it except with your explicit consent, to resolve a specific support issue you have raised, or where required by law.

You can revoke Earshot's access to your Google account at any time, from inside Earshot (Settings → the mailbox → Disconnect) or from your Google account's permissions page. Disconnecting revokes the grant at Google and destroys our copy of the token. Mail already read stays in your CRM until you delete it, which you can do separately.

Who else processes it

Earshot uses a small number of sub-processors, each for one job. None of them receives your data for their own purposes.

WhoFor whatWhat they get
Google Firebase (Authentication, Firestore) Sign-in, and the database your CRM lives in Everything stored in your account
Vercel Running the API and serving this website Requests in transit; nothing is stored there
OpenAI Turning speech into words, when transcription is set to the cloud backend Chunks of meeting audio, for the length of one request. Not retained, not used for training
Anthropic Writing call summaries and contact profiles Transcripts and correspondence, for the length of one request. Not retained beyond the provider's own abuse-monitoring window, not used for training

Data is stored and processed in the United States. If you set both transcription and profile building to run locally, neither OpenAI nor Anthropic receives anything at all.

How it is protected

How long it is kept

For as long as your account exists, because a CRM whose history evaporates is not a CRM. There is no automatic expiry. Everything below is under your control, immediately and without asking us.

Your rights, and the buttons that exercise them

We do not need to ask you to identify yourself for any of this: you are signed in, and the buttons only ever reach your own account.

Children

Earshot is a tool for professional correspondence and is not directed at anyone under 16. We do not knowingly collect information from children.

Changes

If this policy changes in a way that affects what is collected or who sees it, the app will say so before the change applies to you. The date at the top always reflects the current version.