Privacy

A writing assistant has to see what you write. This page is the plain account of what Veda sees and what happens to it. The formal version is the privacy policy.

#When Veda reads your screen

When you press the key — and, with memory always on (the default; switch it to when Veda is used in Memory), the chat or email window in front of you every few seconds while you're at the keyboard (see Memory). Veda doesn't record your screen and doesn't log keystrokes. When you press the key, it reads:

• The text field you're in — its text, and what you've selected.
• The conversation around it — the text above the message box in the same window, so it can reply in context.
• The app's name and the window title — to tell Gmail from Jira inside a browser.

There's one more moment, if learning is on: after Veda puts text in a field, it checks that field for about a minute to see whether you edit its wording before sending. That comparison happens on your computer and is used to learn your style. If usage sharing is on, Veda also sends one number from it — how much of its text you changed — never the text itself.

Password managers and every app you switch off under Which apps Veda remembers are never used for learning either. Private windows and banking sites (unless you turn Banking & finance on) are skipped too, all checked before anything is read.

#Memory, Ask Veda and Brief

With memory on, the conversations Veda sees are kept in an encrypted vault in Veda's folder on your computer, nothing more. They never go to Veda's server. When you ask a question or make a brief, your question and the matching messages (or meetings) go to the AI to write the answer, the same way a rewrite does, and the AI doesn't train on it. Pause, forget by time, delete any item, or wipe the vault from the Memory or Privacy page. Anything that looks like a key, card number or password is replaced with [redacted] before it's remembered or sent with a question. Questions about Veda itself are answered from these help pages and your settings (switches and permissions, never your text).

#Calendars and connections

Apple Calendar is read on your Mac through macOS's own permission. Google Calendar uses Google's sign-in: Veda asks only to see your calendars and events (read-only), the sign-in token is kept on your computer with your other keys, and only the meetings a brief or a question needs go to the AI. Veda's use of information received from Google APIs follows the Google API Services User Data Policy, including the Limited Use requirements: calendar data is used only to show your meetings in Veda and to answer your own questions, and is never sold, used for advertising, or shared with anyone but the AI provider that writes the answer you asked for. Disconnect removes the token and revokes Veda's access at Google.

What Veda does with your meetings, and what leaves your computer:

• The meeting card when a meeting starts (join button, who's in it, the latest from memory, the agenda) and the Meetings page are made on your computer. Nothing is sent.
• Free times in replies. When a reply or draft you asked for is about finding a time, your free times (9 am to 6 pm, Monday to Friday) go to the AI with it. The titles of your other meetings don't.
• People. Emails, full names and time zones from invites are matched to people Veda already remembers, in the encrypted memory file. A calendar never adds new people.
• Meeting notes. Veda listens only between Take notes and Stop — your microphone and the call's sound, on your computer, with no bot in the call. Each minute of audio is turned into text on your computer and thrown away; audio is never saved or sent. When you stop, the transcript and your jottings go to the AI once to write the notes. The transcript and notes stay in the encrypted memory on your computer. Tell people on the call you're taking notes.
• Speaker names. To tell the voices on a call apart, Veda uses small speaker models on your computer. Nothing about a voice is kept after the call — no voiceprint, nothing saved, nothing sent. Names come from the call window (only the name shown as speaking is read; no other window text is kept) or from what was said; the note keeps each voice's name, not its sound. The models come from Veda's own download, not a third party. Speaker model: WeSpeaker ResNet34-LM (VoxCeleb), CC BY 4.0; pyannote segmentation-3.0 (MIT); Silero VAD (MIT). On Windows, speaker matching is calibrated with the AMI Meeting Corpus (University of Edinburgh et al.), CC BY 4.0.
• Notes end with the call. While notes are on, Veda checks the call's window (its title, or a browser's tab titles — never the page) so it can stop a few seconds after you leave the call.
• Screen sharing. The meeting card shows in screen sharing and recordings like any other window. While another meeting is on, the card keeps its notes from memory folded until you click Show.
• Brief and Ask can use double bookings, invites you haven't answered, hours in meetings and past meetings with someone. They go to the AI only inside a brief or an answer you asked for.

#Permissions

On a Mac, setup asks for three: Accessibility (the only one Veda needs to write for you), Microphone and Notifications. The rest are asked for the first time a feature needs them: Call audio when you first take meeting notes and Calendar when you connect Apple Calendar. Veda never asks for Screen Recording. On Windows, Veda shows whether the microphone and notifications are allowed; nothing else needs a permission. Settings → Privacy → Permissions lists them all.

#Where your text goes

On the trial, Veda Free, Veda Pro and Veda Max

Your computer → Veda's server → the AI → back into your field.

Veda's server passes the text to the AI and sends the answer back. It never stores or logs your text or the reply. For each request it records only which account made it, the day, and the request and token counts.

What the AI providers keep

The trial, Veda Free, Veda Pro and Veda Max use OpenAI. OpenAI doesn't train its models on what Veda sends, and Veda doesn't ask it to save anything. Like every business that uses OpenAI's API, OpenAI may keep a request for up to 30 days only to check for abuse, then deletes it. Nothing is stored on Veda's server.

#Your voice

The microphone is on only between your double-tap and the tap that ends it, and the line along the text box shows it's listening the whole time. Your voice is turned into text on your computer — Apple's speech model on a Mac, Whisper on Windows — and the recording is deleted as soon as that's done. It never goes to Veda's server or to the AI; only the text does, to be written up, like a rewrite. This is the same on every plan.

#What's kept, and where

Show folder on Veda's Privacy page opens the folder on your computer.

#Updates

Every minute Veda asks its server whether there's a newer version — that request carries only your app version. Veda downloads the new version and installs it itself, but only if it checks out: on a Mac it must be signed by Veda, on Windows it must match the release exactly. Turn that off with Install updates automatically in Settings → Support, and Veda just tells you instead. The Microsoft Store version asks the Microsoft Store instead, every 10 minutes, and installs what the Store offers.

Every minute Veda also checks for short messages from us, like a new feature or "please update". Like the update check, that request carries only your app version and Mac or Windows, nothing about you: Veda decides on your computer which messages apply to your version and plan. They show under Notifications in the sidebar; turn off their notifications and Home cards in Settings → General, under Messages from Veda.

#Usage data

Veda never stores your text or voice unless you send us a report or feedback. If usage sharing is on, we see which features you use, in which apps, and any errors.

Share anonymous usage data on Veda's Privacy page is on unless you turn it off. It helps us find bugs and see which features work. When it's on, Veda sends short events like "a reply was written in Slack, in an email-type field, in 1.8 s, 40 words in and 62 out" or "voice failed: no speech". Each one includes:

• The feature — rewrite, reply, voice, the edit bar, Read — and whether it worked, failed (with Veda's own error message) or was cancelled.
• The app's name — like Slack or Google Chrome — and the kind of field: chat, email, doc, tracker or general. Never the window title.
• Numbers — how long it took, word counts, and how much of Veda's text you changed afterwards.
• Settings you change — which setting and its new value. Anything you wrote or anything secret (your notes, presets, style, keys) is sent only as "changed".
• About the app — Veda version, Mac or Windows, OS version, trigger key, plan and language setting.
• Messages and the check-in — which of our messages you saw, opened or closed (by number), the answer you pick on the one Home check-in, and the topic of any feedback you send. Never what you wrote.

It never includes the text in the field, your selection, the conversation, what the AI wrote, your voice, your keys or window titles. Turn the switch off and nothing more is sent; events still waiting to be sent are deleted. Usage events are kept for up to 13 months. Because you sign in with Google, they're tied to your device and so to your email.

#Signing in with Google

You sign in with Google in setup's Account step. Veda asks Google only for your name, email address and photo — nothing from your mail, files or calendar. Our server checks Google's answer and keeps your name and email with each computer you sign in on, so your plan and trial follow you (up to 3 computers), and so we know who uses Veda and can write to you about your trial or a problem. The photo stays on your device. Click your name at the foot of the sidebar → Sign out, and the server deletes your name and email for that computer at once; Veda then shows its sign-in page until someone signs in.

#Reporting a bad result

If Veda gets something wrong, click Report last result in the menu-bar or tray panel. The window shows exactly what would be sent — the text Before and After, and your note — and nothing goes until you click Send. Untick Include my text to send only the app, the feature and your note. Reports work even with usage sharing off, because you're the one sending them. The last result is kept in memory only, never on disk.

Feedback. For anything else — an idea, a problem with no single result to point at, something you like — use Give feedback at the foot of the sidebar, in the menu-bar or tray panel, or in Settings → Support. It sends the topic you pick, what you type, where you opened it from (and your answer, if it came from the check-in), and your Veda version and Mac or Windows. Add diagnostics (on unless you untick it) adds settings and states, never your text, names or keys, and See what's sent shows them first. Your email goes with it only if you want a reply. Like reports, feedback works with usage sharing off.

#This website

The website uses Google Analytics only if you click Accept on its cookie notice, and never if your browser sends Do Not Track. Cookie settings in the site footer changes your choice. The app itself sets no cookies.

#Your writing style

The learned profile and the samples it's built from stay on your computer. They leave it only inside requests to the AI: samples during a learning pass, and the profile with each request so the result sounds like you. Neither is kept on Veda's server.

#Starting over

Forget everything learned on Veda's Privacy page deletes every saved sample and resets the learned profile. Your own rules stay — you wrote those. It can't be undone.

To have your account data deleted from our server, email [email protected].

#Payments

Payments go through Paddle, our merchant of record. They handle your card, tax and receipts; we never see your card details. Paddle tells Veda's server about purchases, renewals, cancellations and refunds, so it can start, change or end your plan.