Privacy Policy

This policy describes what Kept stores, what leaves your device, and who else can see it. It is written to match what the product actually does today, including the parts that run on our servers.

The short version. Dictation can run entirely on your Mac. Meetings are different: if you record one and let the app write notes, the transcript is sent to our servers and on to an AI provider. If you are in a team, your teammates can see meetings shared with it.

1. Working offline, and working signed in

The on-device transcription engine runs locally. In that mode your audio is not uploaded, and if you never sign in, nothing is stored on our servers at all.

Signing in changes this, because syncing is the point of an account. Once signed in, the app uploads the items below so they reach your other machines and, where you have chosen to share them, your team.

2. What we store when you are signed in

• Your account: name, email address and which method you signed in with, all provided by Auth0. Also your plan, your settings, and a record of which policies you have agreed to.
• Transcriptions you choose to keep in history.
• Meetings: the transcript, the notes you typed, the AI-written summary, and the follow-ups extracted from it.
• Meeting video and screenshots, only if you turn screen recording on. It is off by default. Video is deleted automatically after the retention period set in Settings.
• Your devices: for each machine you sign in on, its name, operating system, app version, time zone, and which system permissions the app holds. This is what lets support answer “why did my recording come out silent”.
• Usage counts: how many AI requests your account has made, so plans and limits can be applied.
• A truncated network prefix when you accept a policy — the first three parts of an IPv4 address, never the whole one, and never a location.

3. Where your meeting audio and transcripts go

Transcription. With the on-device engine, audio is transcribed on your Mac. If you select a cloud transcription engine (OpenAI or Sarvam) and supply your own API key, your audio goes from your device to that provider. We do not proxy or store it.

Summaries, follow-ups and answers. These do not run on your device. When you ask for a summary, a folder brief, or an answer about a meeting, the relevant transcript text — and, if screen recording is on, extracted screenshots — are sent to our servers and from there to an AI provider on our account. This is a change from earlier versions of this policy, which described AI as running on your own key.

We instruct our providers not to train on this content, and we do not train models on it ourselves.

4. Recording other people

The app records the other side of a call when you turn system audio on. Those people are not our users and have no account here. Telling them the call is being recorded is your responsibility, and in many places it is a legal requirement. See Recording & consent.

The app never starts recording on its own. Meeting detection offers; you decide.

5. Who else can see your data

• Your team. Meetings shared with a team are visible to its members. Each team has a default sharing setting, which you can see and change in Settings, and meetings marked personal are never shared.
• Anyone with a share link. Links you create are unlisted but not secret; whoever holds one can open it until you revoke it.
• Team members who never installed the app, by email. Team digests contain the brief and that person’s own commitments — never a transcript. Every digest carries an unsubscribe link.
• Our staff, where an administrator needs to investigate a support issue or abuse. Administrative access is restricted and blocking actions are logged.

6. Analytics and crash reports

Usage analytics are off unless you turn them on. When on, the app sends event names, app version and platform — never transcribed text, audio, or API keys.

Crash reports are on by default and can be turned off in Settings. A report contains a stack trace and version strings.

The app checks for updates using only the app version and platform.

7. Your API keys

Keys you enter for OpenAI, Sarvam, Anthropic or ElevenLabs are held in the macOS Keychain on your device. They are never uploaded. Our servers record only whether a key exists, so support and billing can tell which provider you are using.

8. Where it is stored, and who processes it

Data is stored in Amazon Web Services and MongoDB Atlas in the Mumbai (ap-south-1) region. We use these processors:

• Auth0 (Okta) — sign-in and account identity
• Amazon Web Services — hosting, file storage, and email delivery
• MongoDB Atlas — the database
• Anthropic and Amazon Bedrock — summaries, briefs and answers
• Stripe — payments, if and when you buy a plan
• OpenAI, Sarvam or ElevenLabs — only if you supply your own key

9. How long we keep it

• Meetings, transcripts and notes are kept until you delete them.
• Meeting video is deleted after the retention period you set. The default is 14 days.
• Deleted items are retained briefly before removal so that an accidental deletion can be undone.
• Deleting your account ends access immediately and erases your data after 30 days. Within that window you can restore it from the app. After it, it is gone.
• Records of which policies you accepted are kept after the account closes, because that is the only period in which the question is usually asked.

10. Your choices

• Export everything from Settings, at any time, including while your account is blocked.
• Delete your account from Settings, with 30 days to change your mind.
• Turn analytics, crash reports and email digests off.
• Sign out any device from Settings.
• Ask us for a copy or a deletion by email; we will do it.

11. Children

Kept is not intended for anyone under 16, and we do not knowingly collect their data.

12. Changes

When a policy changes materially we ask you to agree to it again in the app, and we record which version you accepted and when.

13. Contact

support@aithinkers.com. We read every email.