Privacy

Your account: name, email address, and an avatar if your sign-in provider offers one. If you sign in with Google, we also store the tokens Google issues for that connection so it keeps working. If you set a password, we store a hash of it, never the password itself.

Your workspace content: projects, tasks, comments, labels, activity history, meeting transcripts and summaries, transcripts you approve, and the reports computed from them. This content belongs to your team, we process it only to run the product.

Technical basics: session cookies (sign-in only, no advertising trackers), standard server logs and error reports to our own team, kept briefly for security and debugging and stripped of what you typed, and, if you turn push notifications on, a registration token identifying that browser or device.

Website analytics: on this website we count page views through Vercel Analytics, which sets no cookies, builds no profile and follows nobody across other sites. It records the page, the referrer and coarse details like country and device type, never your name, your email or anything inside your workspace. The Android and iOS apps do not run it at all.

When you use voice capture, the audio is transcribed to text so you can review and confirm the tasks we extract from it. By default the recording is sent to our transcription provider, Deepgram, to be turned into text. Under our agreement with them it is transcribed and not kept afterwards. We keep the transcript, not a stored voice-capture recording. Discard a capture and it’s marked discarded; delete your account or your workspace and it’s gone.

Each transcription request also carries a short vocabulary list built from your workspace: teammates’ names, project names, client names and label names. It is what makes local names come out spelled right, and it is sent whether or not the recording mentions any of them.

If a device can’t record for us, we fall back to the speech recognition built into your browser. That fallback is not private to your device: Chrome and other Chromium browsers stream the microphone to Google’s speech service to do the recognising. This affects the website only. The Android and iOS apps have no such browser feature, so there the Deepgram path is the only one.

Meetings are an opt-in feature. When you record one, the audio is captured (on your device, or by a clearly-named notetaker that joins the online call as a visible participant so everyone can see it) and uploaded to our storage. Please tell everyone in the room they’re being recorded before you start, it’s the law and it’s good manners.

Meeting audio leaves our infrastructure. The recording is stored with Supabase, our database and file host, and Deepgram fetches it from there to transcribe it. The transcript is then sent to Anthropic to write the summary and propose action items. If you use the notetaker bot instead of recording on your own device, Recall.ai captures the call and holds the audio briefly before we copy it across.

Recordings are private to whoever made them, admins included, until they share the notes or link the meeting to a project. You can delete the audio at any point and keep just the notes, which removes the recording from storage. Action items stay proposals until a person confirms them.

Transcripts, typed quick-adds and weekly activity summaries are processed server-side through Anthropic’s API to extract task proposals and write your briefing. To attribute work to the right person, those requests also include the names of everyone in the workspace, whether or not the content mentions them. We do not send members’ email addresses. Anything you type or say is sent as you wrote it, so an address inside a note or a recording goes with it. Under Anthropic’s commercial terms this data is not used to train their models. AI credentials never reach your browser, and the AI never creates or changes work, a person always confirms first.

If you connect a Google Calendar (an optional feature for workspaces with the meeting notetaker), Alpha Workspace asks Google for read-only access to your calendar events, the calendar.events.readonly scope. We use it for one purpose: to list your upcoming events that carry a video-call link, so you can choose which calls the notetaker joins. For each such event we keep the event id, title, start and end time, the join link, and whether it was cancelled. We do not read or store attendee lists, descriptions, or any event without a call link, and we never write to your calendar.

The refresh token Google issues is handed, in the same request, to Recall.ai, the service that runs the notetaker and keeps your calendar in sync; Alpha Workspace does not store it. Disconnecting the calendar under Account revokes that access at Recall.ai and deletes the events we hold; recorded meetings stay because they are yours, not the calendar’s. Nobody else in your workspace can see your calendar events, and no workspace owner or administrator can connect a calendar on your behalf.

Alpha Workspace’s use and transfer of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements. Google user data is never sold, never used for advertising, and never used to train AI models. Humans read it only with your consent, for support or security, or as the law requires.

We use a short list of operators, each doing one job:

• Supabase, the database and the file storage behind attachments and meeting audio.
• Vercel, application hosting and the website page-view counts described above.
• Resend, email delivery: sign-in links, invites, nudges and meeting notes.
• Anthropic, the summaries and task extraction described above.
• Deepgram, speech-to-text for voice captures and meetings.
• Recall.ai, the optional meeting notetaker bot, only if your workspace has it switched on.
• Paystack, payments. Card details go directly to Paystack and never touch Alpha’s servers; Paystack stores the card and bills the subscription. The payment events Paystack sends us afterwards are recorded for audit as event names and references, without card details.
• Google, in three separate roles: sign-in, if you use it; hosting the avatar image that comes with a Google account, which means your browser fetches it from Google whenever you view that person’s avatar; and Firebase Cloud Messaging, which carries push notifications to the Android and iOS apps. Push to an iPhone reaches Apple through Firebase, so the device token and the text of the notification pass through Google on both platforms.
• Your browser’s push service (Google, Mozilla or Apple, depending on the browser), if you enable push on the website. The contents are encrypted, but the service necessarily sees that a message was delivered to your browser.

Some of these process data on servers outside South Africa (for example the United States); we only share what each needs to do its job. We don’t sell personal information. Ever.

You can export your data as JSON at any time, from Account, Your data, Export. The file contains your profile, your workspace memberships, tasks assigned to you, tasks you created, your comments, your voice captures, your notifications and your private tasks.

It does not currently include everything: meetings and their transcripts, attachments, time entries, scorecard entries and activity history are not in the automatic export. Ask us at joseph@underbridges.co.za and we will put those together for you.

You can delete your account, and an owner can delete a whole workspace, from inside the product. Deleting your account removes your profile, your sign-in credentials and sessions, your private tasks, your voice captures, your logged time, your notification preferences and history, your push registrations, and the meetings you recorded together with their transcripts and summaries. A workspace you solely own, with nobody else in it, goes too, and everything in it with it.

Two things deliberately survive. Work you authored in a workspace that other people still use stays with that workspace, with your name taken off it, so the team keeps its record of its own projects without keeping you in it. And the activity log stays as de-identified history, because the scorecards and the weekly narrative are computed from it; after deletion those entries point at nobody.

Billing records are kept for as long as South African tax and company law requires us to keep them.

The full deletion process is written out here, including what to do if you can no longer sign in.

The database, the files and the servers that run the app are in London, in the United Kingdom. Some of the providers listed above process data in other countries, including the United States, to do their part of the service, and only that part.

Why we process it: to provide the service your team signed up for, to keep it secure and working, and, for the features you switch on yourself (voice capture, meeting recording, a connected calendar, WhatsApp), because you asked us to. You can switch any of those off, and that stops the processing from then on.

You can access and correct your information in the app, export it, and delete it, as above. Notifications are tunable per type and channel (Account, Notifications); external channels are outbound nudges only and never collect replies.

Questions or complaints: raise them with your workspace owner, or write to us at joseph@underbridges.co.za. Under the GDPR and the UK GDPR you can also ask us to restrict or stop a use of your information. You may approach your data protection authority: the Information Regulator in South Africa (inforegulator.org.za), the ICO in the United Kingdom (ico.org.uk), or your local authority in the European Union.

Creating an account is your consent to process the information above for the purpose of running Alpha Workspace, nothing broader. This page changes only with notice in the product.