Museda privacy notice
Museda is a private personal AI. It can only be used by invitation from its owner, who runs it in their own cloud account.
What data it keeps
- Your conversation with Museda, the files you send it and the notes it writes to remember you.
- Your goals, reminders and, if you turn them on, the signals you share (such as your location).
- To identify you: the anonymous identifier provided by “Sign in with Apple” (and your email, if Apple shares it).
- If you connect services (such as Google or Outlook) or save passwords, they are stored encrypted and Museda only uses them when you ask it to.
What it is used for
Only so Museda can answer and help you. To generate responses, your messages are processed with a language model: GLM, by Z.ai, by default, so they are sent to Z.ai, which processes them in Singapore and, under its terms, does not store them or use them to train its models. If you prefer Claude, by Anthropic, choose it in Settings → Model; with Claude, what Museda does on its own (its check-ins, the diary) uses its Sonnet model. With either one, web searches and reading images and PDFs are done with Claude. Each goal you follow has its own chat in the app with an AI-made image: to create it, Claude describes in a few words what to draw from the title, and only that description is sent to Z.ai (its image model). The same goes if you ask the AI for Museda's photo (Settings → Personality): only a short description of how you want it to look goes to Z.ai. If you ask for a new image (a poster, an invitation), the description written by the AI goes to Z.ai to draw it. If you ask it to edit a photo (change something, remove the background, enlarge it), it is processed with Stability AI models on Amazon Bedrock, inside Amazon Web Services, without being sent to Stability AI. If you give it a YouTube link, Gemini, by Google, only receives that link and the instruction to transcribe and describe the video (not what you asked); Google does not use that data to train its models. The videos and audio files you send never leave Amazon Web Services: the audio is transcribed with Amazon Transcribe, which does not keep it, and the language model sees a few frames of the video. If you ask for something that needs a browser, Museda opens the pages in an isolated one on Amazon Web Services. It only uses your account on sites where you yourself signed in from the app, in a live view of that browser (Museda never sees your password); those sessions are kept in your browser profile and you can delete them in Settings → Museda's browser → "Forget all sessions". It never submits a form (a purchase, a booking, a message) unless you approve it in the app (or in your Telegram, if you connected it). Before each click, JEV, a model by TypeSafe AI, estimates whether the action commits you to anything; for that it receives the page address, the text near the button, the names of the filled-in fields (not what you typed in them) and your latest messages in that chat. It can only make Museda ask for your approval more often, never less. Your data is not sold, not used for advertising and not shared with anyone else.
Who can see it
Each person has a separate space: other users can't see your information and the app doesn't show your conversation to the owner. On top of that, your data is stored encrypted with a key of its own for your account, which only the Museda service can open while it works for you: whoever administers the cloud can't read it directly. There is a single emergency access, protected with two-step verification, and every use of it is logged and triggers an alert right away. Even so, the data lives in the owner's cloud account, which controls the service and could change it; that would be logged too.
Apple Health data
If you turn on «Compartir mi salud» (Share my health) in the app, Museda reads a daily summary from the Health app (HealthKit): steps, distance, exercise, active energy, sleep, resting heart rate, heart rate variability, weight, mindful minutes and workouts. You choose what it can read in iPhone Settings → Health, and you can revoke access at any time. Museda never writes to Health.
iPhone calendar, reminders and contacts
If you turn on «Calendario y recordatorios» (Calendar and reminders) in the app's Settings, your iPhone sends Museda your events (from two weeks back to two months ahead, from every account on the iPhone), your lists and your pending reminders plus those completed in the last week; if you turn on «Contactos» (Contacts), the names, phone numbers, emails, addresses, companies and birthdays of your contacts (no photos or notes). They are stored encrypted, used only to help you, and the AI reads them when useful, so they also go through the AI model you use. Museda only adds a reminder, marks one as done or adds an event when you ask (events, with your approval); it never deletes anything on your iPhone. When you turn either one off, what it had sent is deleted from the server. You can also revoke access in iPhone Settings → Museda.
This data is only used to support you (giving context, noticing changes in your sleep or activity and recording them in your diary). It is not used for advertising or to train AI models, it is not sold and it is not shared with third parties, except for processing it with the language model to answer you.
Google data
If you connect your Google account, Museda uses Gmail, Google Calendar and Google Drive only for what you ask: reading, searching and organizing your email, viewing your calendar and files and, only with your approval each time, sending emails, creating events or uploading files. When you connect your email (Gmail or Outlook), Museda checks each new email as it arrives: the sender, the subject and the first lines go through JEV, a model by TypeSafe AI (or, if it doesn't respond, through Claude, by Anthropic), to decide whether it is worth telling you right away; TypeSafe receives nothing else from the email. If it is, Museda reads it and writes to you. This data is not used for advertising or to train AI models, it is not sold or transferred to third parties (except for processing it with the AI models described here, to answer you and alert you), and it is not read by humans, except with your permission, for security purposes or to comply with the law.
Museda's use and transfer to any other app of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements.
You can disconnect Google at any time from Settings in the app (or from your Google account, under “Third-party apps with account access”): the saved permissions are deleted.
Artifacts and public links
Museda can make you artifacts: pages (a dashboard, a tracker) that you open in the app and that keep what you enter in them. They are shown from a separate domain, isolated from your account, and can't send your data to other sites. If you share one, anyone with the link can open it: as a snapshot (the page and its data at that moment) or live (with up-to-date data from its sources, such as your calendar or email; the app tells you which ones before publishing). Whatever people who open it change stays in their browser. You can turn the link off at any time.
If you connect Telegram (Settings → Telegram), what you write to Museda there and its replies go through Telegram's servers, which store them in its cloud: chats with bots are not end-to-end encrypted, and Telegram's own terms apply to them too. Voice notes you send through Telegram are transcribed with Amazon Transcribe, in the same Amazon Web Services account, without keeping the audio. You can disconnect it whenever you want; whatever is already in Telegram is deleted from Telegram.
Deleting your data
Ask the owner to delete your account: your conversation, your files and everything else are permanently removed. Location and health records are deleted automatically after 90 days.