Privacy Policy

Last updated: October 8, 2026

The short version

You talk with Mazalot out loud. To make that work we collect your name and phone number, your birth details, the words of your conversations, and notes that AI builds from those conversations. Right after you first sign in, and before your first reading or call, the App asks you to agree that outside companies process your readings, calls and meditations; that screen links to this policy, which names them. Your voice is processed live by our voice provider. Your calls are recorded and kept for 30 days so we can check quality and fix problems, then deleted. The recordings stay with our voice provider; our own servers never store the audio. The written transcript is kept with your account. We send your data to a small set of providers who run parts of the App for us, and we name every one of them below. If you link with another user, you share names and birth charts, and each of your astrologers may be guided by a single label from a fixed list, drawn from the other person's own conversations, to help you understand each other; an astrologer never repeats, quotes, summarizes or attributes anything the other person said, and your words are never shared. We do not sell your information, we do not share it for advertising, we show no ads, and the website sets no cookies. A few named staff can read transcripts, and listen to a call recording while it exists, for support, safety and quality, and every opening and every listen is logged. You can delete your account in the App or from mazalot.ai/delete-account; deletion becomes permanent after thirty days.

1. Who is responsible for your data

Mazalot Inc., a California corporation, is the business that decides how and why your personal information is used. Mazalot Inc., 11631 Victory Blvd Ste 201, North Hollywood, CA 91606, United States. Email: compliance@mazalot.ai.

2. Who this policy covers

This policy covers the Mazalot mobile app for iOS and Android and the mazalot.ai website (together, the "App"). The App is offered in the United States only and to adults only. If you use it from elsewhere, your data is still processed in the United States as this policy describes. This policy is written in the shape of a California notice at collection because that is the clearest way to tell you what we do: section 3 lists each category we collect, where it comes from, why, and how long we keep it; section 11 lists who receives it; section 12 confirms that we sell and share none of it. The App links to this policy on the sign-in screen, in Settings and on the subscription screen.

3. What we collect, where it comes from, why, and how long we keep it

"Life of the account" means until your account is deleted, after the thirty-day grace period described in section 15.

Identifiers

What: your name; your mobile phone number as you typed it and in standard international form; the country your number belongs to; an internal account identifier; an internal sign-in address that our sign-in system builds from your phone number (it never receives mail); the time you accepted our Terms and this policy, and the version label.
Source: you, at sign-in. Why: signing you in with a one-time code; addressing you by name; choosing the right regional crisis resources; matching an invitation a friend sent to your number. Kept: life of the account.

Birth data and your birth chart

What: birth date; birth time, or a note that it is unknown; birth place as you typed it; the place's coordinates and time zone; the computed birth chart, written interpretations of it and of individual placements, and an image of the chart wheel labeled with your name.
Source: you, during setup; the coordinates from our place-lookup provider; the chart from our own calculation, done with the Swiss Ephemeris library. Why: your birth chart, daily reading, synastry, live planetary transits, readings, cards and meditations; the birthday gift when we run it. Kept: life of the account (section 15 explains how the chart image is removed at deletion).

Your agreements

What: that you agreed on the permission screen before your first reading or call (section 6), when, and whether you later withdrew it; for each link with another user, who invited whom, when each of you agreed, and when the link ended; and, if you tapped "Not now" on the card described in section 8.6, when you did, so that the card is not shown again too soon.
Source: you. Why: doing only what you agreed to, and being able to show it. Kept: life of the account; the Confidant part of a link is deleted when the link ends (section 8).

Conversation transcripts and call recordings

What: the words you and the companion say on each call, as text, with the call's start, duration, language, how it ended, the app platform and version used, and a reference to the voice provider's copy; and the call recording, the sound of you and the companion on the call, which our voice provider makes and keeps for us.
Source: transcribed live, and recorded, by our voice provider; the transcript is delivered to us after the call. Why: the conversation notes in section 7; support and safety review and the weekly de-identified analysis (section 9); reconnecting a dropped call; for the recording, checking quality and investigating problems (section 9). Kept: the transcript, life of the account; ElevenLabs keeps its copy of the written transcript until we ask it to delete it when your account is deleted (section 6). The call recording, kept by our voice provider for 30 days and then deleted automatically; if your account is deleted before then, it is deleted at once with the rest of the account (section 6). Mazalot's own servers never store the audio; when a staff member listens to a recording (section 9), our staff dashboard plays it from our voice provider without saving a copy.

Conversation notes ("memory")

What: what AI extracts after each call: topics, decisions, life events, emotional themes, unresolved questions, the people in your life, personality patterns the AI notices, moods, chart placements that resonated, advice given, short summaries, private aims, and a flag if the call showed signs of crisis; a label saying whether a call's notes are what you said, what you implied, or what the AI inferred; a knowledge base about you, open questions to return to, a written portrait of you, the "Moments" the App collects, and, when enabled, numeric representations (embeddings) of notes.
Source: generated by AI from your transcripts. Why: shaping later conversations, readings, cards and meditations. Kept: life of the account.

Other people

What: for people you add to your list: name, relationship, gender, birth date, time and place, coordinates, time zone, their chart, an AI reading about the relationship and a short periodic line about it. For people you mention on calls: name or nickname, how they relate to you, facts you mentioned, how often, the tone of the relationship as heard from you, and a flag if you spoke of them with fear. If you invite someone by phone: their phone number and whether it matched an account.
Source: you; AI, from your transcripts. Why: charts and synastry readings; keeping track of who is who; linking two users who both agree. Kept: until you remove the person, or life of the account; see section 8 for invitation numbers.

Linking and the Confidant

What: while you are linked with another user: their name, the birth date, time and place from their account and their chart, a reading of the two of you; and, for the Confidant, labels for relationship needs you seem to have in each bond, each chosen from a fixed list of 17 (for example "undivided attention" or "space to process"), each with a stage, a strength score and the date it was last noticed. Each call adds or updates at most one label per bond, and only the strongest current label for a bond can reach the other person's astrologer. The labels have no free text.
Source: the other user's account; AI, from your own transcripts (for your label) and from the other person's transcripts (for theirs). Why: synastry from accurate charts; helping each of you understand the other (section 8). Kept: while the link lasts. Unlinking deletes the Confidant labels on both sides at once and removes from your entry what came only from the other person's account (section 8).

Purchases and minutes

What: subscription status and plan, store transaction identifiers, a ledger of every minute granted, spent and given back, refunds and reversals, referral and creator codes you created or redeemed, and gifts granted. When you buy, we pass your internal account identifier to Apple or Google Play so the purchase is linked to your account. We never see your card number.
Source: Apple, Google Play and our own systems. Why: giving you the minutes you paid for; applying the Terms; reversing refunded purchases; support. Kept: life of the account.

Device, usage and experience signals

What: platform (iOS or Android), app version, your device's current time zone, your push notification token, the network address (IP) of requests, and product events such as setup steps completed, calls started and ended, screens shown and which paywall you saw. The age range check at your first sign-in is recorded as a product event with the age range your store reported (its lower and upper bound, if any) and the result, and your account is marked as checked so the check does not run again. From how you use the App (how calls end, call length, return visits, meditations finished, errors, running out of minutes) we compute a simple experience label for your account: good, poor, or neither.
Source: your device; your app store; our servers. Why: running the App; keeping you signed in; sending notifications you enabled; enforcing the adults-only rule; preventing abuse; understanding how the App is used; deciding when to show the "How's it going?" feedback card. Staff can see the label. It is never used for advertising or to make decisions about your account. Kept: push token until sign-out or deletion; phone numbers and IP addresses in our abuse-prevention table for about two days, sometimes longer before cleanup; the experience label for the life of the account; product events indefinitely, with your account identifier removed at deletion.

Feedback and store reviews

What: stars and notes you send through "Send feedback" or the "How's it going?" card, including reports of a bad AI answer, with your experience label; our staff's replies. Reviews you post publicly on the App Store or Google Play (reviewer nickname, territory, device, text and stars), which we collect from the stores daily, and our best guess, from timing, app version and region, at which account posted them.
Source: you; Apple and Google Play. Why: answering you, fixing problems, replying publicly to store reviews. A daily summary of low-star reviews and notes, and of notes sent while your experience label was poor, quoting their words but no names, is emailed to our company inbox. Kept: life of the account for your notes; store reviews remain after deletion with the account guess removed.

Meditation and wake-up audio

What: meditation scripts written for you, the generated audio, a short clip of the companion's synthetic voice saying your first name, your favorites and feedback on meditations, your voice choice, and your wake-up time.
Source: generated by us; you. Why: playing meditations and wake-up readings. Kept: life of the account; section 15 explains how audio files are removed at deletion.

Notification settings

What: whether the daily reading notification, check-in rings and the wake-up alarm are on, and your times. Source: you. Why: sending only what you asked for. Kept: life of the account.

Crisis records

What: a referral record with the date, the kind of statement our rules matched and a two-letter region, with no account identifier and no words; and a product event noting that a call's notes were flagged for crisis, which carries your account identifier until deletion.
Source: our crisis detector and after-call processing. Why: responding safely, and the annual reports state law requires. Kept: referral records indefinitely in de-identified form; the event as product events are kept (identifier removed at deletion).

Website visits

What: nothing is stored by us. Vercel, which hosts the site, and Google Fonts receive your IP address when they serve the pages and fonts. If you email us, for example to delete an account from mazalot.ai/delete-account, we keep your email and our reply as section 14 describes. Kept: page visits are not kept by us.

During setup the App asks how you describe yourself (Woman, Man or Rather not say). That answer stays on your phone during setup; it is not sent to us or stored with your account. If we start using it to shape how the companion speaks to you, we will add it to this section first. We do not ask for your email address, your precise device location, photographs, contacts or payment card details. The App asks for the gender of people you add to your list, because relationship astrology uses it.

4. Where the information comes from

Almost everything comes from you, either typed into the App or spoken on a call. The rest is generated by our systems and AI providers from what you gave us, comes from another user you chose to link with, or comes from Apple and Google Play about your purchases, your public store reviews and, at your first sign-in, the age range your store reports. We do not buy information about you from data brokers and we do not collect information about you from social networks.

5. What we use it for

To sign you in and keep your account secure, including checking the age range your store reports.
To calculate your birth chart and generate your conversations, daily reading, readings, cards, meditations and wake-up readings.
To keep the conversation notes described in section 7.
To keep track of the people in your life and, if you both agree, to link you with another user and run the Confidant (section 8).
To give you the minutes you paid for and to apply the rules of the subscription, packs, trial, gifts and referrals.
To send you the notifications you turned on and to ring the alarm you set.
To respond when a call suggests you may be in crisis, and to report de-identified counts of those referrals where the law requires.
To support you, answer your feedback, review reports of bad AI answers, investigate abuse, keep the App safe, and enforce our Terms.
To understand how the App is used and improve it, including the experience label and the de-identified weekly analysis in section 9.
To comply with the law and to protect our rights and the rights of others.

We do not use your information to make decisions that have legal or similarly significant effects on you, and we do not profile you for advertising.

6. Your voice, the permission screen, and the AI notice

6.1 The permission screen

Right after you first sign in, before you enter your birth details, the App shows a screen titled "Before your first reading". It tells you, in plain words, that Mazalot uses trusted AI partners to create your readings and voice calls, that we share only what they need and they never use it to train their AI, and that calls are recorded for quality and deleted after 30 days. The screen does not name the companies: its "See the full details" link opens this section of this policy, and the companies are named, with what each receives and how long each keeps it, in sections 10 and 11 (Anthropic for the AI that writes your readings and the companion's replies, ElevenLabs for the voice and, when the App's search of your notes is switched on, OpenAI). What you share with the companion can include your health and feelings; the sound of your call goes to ElevenLabs as you speak, and the words are kept as a written transcript (section 3). At the start of each call the call screen shows the line "Recorded for quality, deleted after 30 days." small and quiet for the first 5 seconds; the line then fades out and does not stay on screen. Our own servers never store the audio; the recording stays with our voice provider. You tap "I agree" before your first reading or call, or "Not now", which leaves the App unable to make readings or calls for you until you agree; if your account existed before this screen, the App shows it the next time you open the App. Until you agree, the App makes no readings, calls, meditations or wake-up readings for you, so our server sends none of your birth details or conversations to these companies for them. You can withdraw that agreement in Settings; the App then stops making readings, calls, meditations and wake-up readings for you until you agree again. Withdrawing does not delete what was already processed; deleting your account does (section 15).

6.2 How a call is processed

When you start a call, the App opens a live audio connection from your phone to our voice provider, ElevenLabs, over a real-time streaming service (LiveKit) built into their software. ElevenLabs turns your speech into text as you speak and sends the text to our server; our server gets the companion's reply from our language model provider, Anthropic, and sends it back as text; ElevenLabs turns it into the speech you hear. The audio flows through ElevenLabs. During a call, Mazalot's own servers never receive the audio, and they never store a recording of your voice.

ElevenLabs records each call. The recording is kept for 30 days so that we can check quality and investigate problems, and then ElevenLabs deletes it automatically. ElevenLabs also keeps the written transcript of each conversation, under its agreement with us, until we ask it to delete it. When your account is deleted we ask ElevenLabs to delete every conversation of yours, including any recording still kept, right away; if that request fails, we retry every hour until ElevenLabs confirms it is done. Section 9 explains who at Mazalot can listen to a recording.

At the start of every call, our server composes a written briefing so the companion can talk with you as itself, and your phone passes it to ElevenLabs: your name, your chart, facts and aims drawn from your conversation notes, up to ten people in your life with their role and the tone you have used about them, your region, upcoming events you mentioned, a recent meditation, current transits and your schedule, and, while you are linked with someone and the Confidant is on, a line of guidance based on that person's Confidant label (section 8). That briefing is personal information and is covered by the same provider agreement.

6.3 No voiceprint

We do not create a voiceprint. We do not use your voice to identify you, to authenticate you, or to build any template that could recognize you by your voice, and we have no feature that clones your voice. If we ever add a feature that would create a biometric identifier from your voice, we will ask for your separate written consent first and publish a retention and destruction schedule, as the biometric privacy laws of Illinois, Texas and Washington require.

6.4 The AI notice

At the start of every call the call screen shows the line "You're talking with an AI astrologer. This is not therapy, medical care, or a crisis service." small and quiet for the first 5 seconds; the line then fades out and does not stay on screen. On a call that reaches three hours, the line fades in again for 5 seconds, and again every three hours after that.

7. Conversation notes ("memory")

During a call, the companion's replies are written live by Anthropic's models, which receive each turn with the conversation so far and your briefing. During a call the server forms short-lived impressions of your state; they stay in its working memory for that call, and only short labels of them, with no account identifier, are kept in our product events (section 18).

After each call in which you said at least five words, Anthropic models read the transcript. A small screening model first sorts the call by how much it contains; then the main job writes the notes listed in section 3. Several smaller jobs run alongside: one lists upcoming events you mentioned, one links threads across calls, one updates the private map of people in your life, one keeps a knowledge base, one tracks open questions, one writes a portrait of you, one collects "Moments", which pair something the companion said earlier with something you said later, and, only while you are linked, have agreed and have an active subscription, one picks your Confidant label for each link (section 8). When embeddings are switched on, OpenAI also receives note text, and, during a call, some of what you just said, so that it can be turned into numbers that help the App find related notes.

Notes include things you never said out loud. The AI infers moods, patterns and relationship dynamics, and the App labels each call's notes as told, implied or inferred.

What you can and cannot do with notes today: with an active subscription you can see the Moments the App has collected. You can remove a person from your list, which deletes their entry and the record kept about them; call summaries that mention them remain. You cannot yet view, edit or delete individual notes, and you cannot mark a single call as one to leave out. The way to erase all notes is to delete your account. If we add these controls, we will update this policy.

A call in which signs of crisis were found is still transcribed and its notes are still stored, with a crisis flag; the flag stops that content from being reused in cards, the Confidant and the smaller jobs above.

8. Information about other people, linking, and the Confidant

Section 3 lists what we keep about other people. Here is how it is used.

8.1 People you add by hand

You can add a person to your list without that person's permission, and the App never contacts them. The entry is the user's own private note: no other user sees it and we build no account and no marketing profile for that person. The name, relationship, gender and chart of each person you add are sent to Anthropic to write a synastry reading about the relationship and a periodic line about it. The person who adds an entry is responsible for having the right to enter those details. Deleting the entry removes it, its chart and reading, the notes linked to it about people you mentioned, and any Confidant information about that bond.

8.2 People you mention in conversation

The notes the App keeps about a person you mention exist only inside your account, to serve you. They are kept to what helps your own conversations: a name or nickname, the relationship, facts you mentioned and the tone you used. They are never shown to any other user, including the person they are about, and they are deleted with the entry they belong to or with your account. Each call briefing sent to ElevenLabs includes up to ten of the people in your world, with their role and a note on the tone you have used about them. The App computes a chart only for people in your list, from the details in their entry.

8.3 Invitations

If you invite someone to link, you enter their phone number, and we store it to match the invitation to an existing account or to the account they later create. A number on an invitation you cancel before it is used is deleted with the invitation; a number on an invitation that expired, was declined or was accepted stays until you remove that person or delete your account. Invitation messages are sent from your own phone, through your messaging app, WhatsApp or the share sheet. Our answer to your invitation never tells you whether that number already has an account.

8.4 Linking and the Confidant

Linking is one agreement between two users that covers two things: sharing charts and the Confidant. Both people see the same explanation before they tap: "Linking shares your name and birth chart with each other. It also lets each of your astrologers gently use what each of you shares in your own conversations, to help you understand each other better. Your astrologer never repeats, quotes or summarizes what the other person said, and never tells you it came from them. Either of you can unlink anytime, and it stops for both of you right away." The link, and both of its parts, starts only when the invited person taps "Agree and link". While both agree, each person's astrologer may be guided by the other person's needs, but it never repeats, quotes, summarizes or attributes anything the other person said; what guides it is a single category from a fixed list, never text.

What moves between your accounts, and when. When the link starts, each of you receives the other's name and birth chart, and the birth date, time and place on each account are copied into the other person's entry, so that both synastry readings use accurate charts. A reading of the two of you is then written by Anthropic.
The Confidant. After each of your calls, while you are linked, have agreed and have an active subscription, an Anthropic model reads your own transcript and may pick, for each link, at most one label from a fixed list of 17 universal relationship needs, with a stage. Before it picks, it screens out anything private, specific or embarrassing. That label, with its stage, a strength score and the date, is stored in your account, and nothing else is; over many calls your account can hold several such labels for one link. A rule in our database lets the other person's astrologer read only the strongest current label for your link, and only while both of your agreements are on; a separate, restricted database connection is the only way the label is read. When the other person next calls, their astrologer may receive a line of guidance based on the label, written by Anthropic and starting from the other person's own chart and notes. It never receives your words, so it has nothing of yours to repeat, quote or summarize, and it is told never to say or imply where the guidance came from.
What never moves. No transcript, quote, summary, word, name of anyone else, or detail of what you said ever reaches the other person or their astrologer, and the database has no place to store such text for the Confidant. Nothing from a call flagged for crisis is used.

8.5 Unlinking

Either person can unlink at any time. Unlinking ends the chart sharing and the Confidant for both people at once. In the same step we delete both people's Confidant labels for that link and both directions of exposure, so neither astrologer can read anything from the other any more, and the reading of the two of you is cleared. Each of you keeps the other as a regular, unlinked entry. Whatever you typed in yourself for that person stays. Anything that came only from the other person's account, such as their exact birth time and place, is removed, and the entry goes back to what you had before; if you never typed anything for that person, because they joined through your invitation, the entry keeps only their name and birth date. The App then writes a new reading of that person from what remains. The record that the link existed, when it started and when it ended stays with your account (section 3).

8.6 People who were already linked

If you were linked before the Confidant became part of linking, the App shows you the same explanation on a card with the buttons "Agree" and "Not now". The Confidant does not run for you until you agree. If you tap "Not now", the link stays, your charts stay shared, the Confidant stays off for you, and the App may show the card again, no more than once every thirty days. You can unlink at any time from that person's entry (section 8.5).

8.7 If you delete your account while linked

Deleting your account ends every link, and the same rule applies as when you unlink (section 8.5). Your chart sharing is deleted, and the Confidant records for each pair you were linked in are deleted in both directions. The other person's entry for you loses its link to your account, the reading of the two of you, and whatever came only from your account, such as your exact birth time and place. It keeps whatever that person typed in themselves; if they never typed anything for you, because you joined through their invitation, it keeps only your name and birth date. Their entry stays in their list as a regular, unlinked entry. You can ask us to remove what remains, as section 8.8 explains.

8.8 If someone entered information about you

If another user has entered information about you and you want it removed, write to compliance@mazalot.ai with enough detail for us to find it, and we will take reasonable steps to locate and delete it.

9. Staff access to transcripts and call recordings, and the weekly de-identified analysis

Staff access. A small number of named Mazalot staff can open the full transcript of a call and its notes. They may do so only to support you, to investigate abuse or a safety concern, or to check the quality of the companion. To open a transcript the staff member must sign in to our staff dashboard with their own email and a one-time code, request the specific conversation, and type a confirmation; access then lasts fifteen minutes. The dashboard writes a permanent audit record of which staff member opened which conversation, for which account, and when. That record is kept indefinitely, including after your account is deleted, so that access can always be reviewed. A few named staff can also listen to the recording of a call while it still exists, which is at most 30 days after the call, for the same reasons and through the same request and typed confirmation. Before any audio plays, the dashboard writes the same kind of permanent audit record: which staff member listened to which conversation, for which account, when, and from which network address. That record is kept the same way. The recording plays from our voice provider through the staff dashboard, which does not save a copy. Staff sign-ins are logged with the network address and rough location, and each one triggers an alert email to the company. Staff can also read feedback you send and reply to it.

Weekly analysis. When we switch this feature on, once a week our server takes a sample of up to eighty recent transcripts, cuts each to its first four thousand characters, and scrubs it: your name and the names of people in your list and people you mentioned, birthplaces, email addresses, links, handles, dates, times, phone numbers and long numbers are replaced with generic tags. The scrubbed text, without any account identifier, is sent to Anthropic, which returns a list of themes and how often each appeared. If the answer contains anything that looks like a name, place, email or long number, it is thrown away. We keep only the week, the sample size, the model used and the themes with counts. Accounts in the deletion grace period are left out. The scrubbing is automatic and not perfect: nicknames we have not recorded, employers and other details can survive it. We do not attempt to re-identify de-identified data, and we do not allow anyone we share it with to try.

10. AI providers and training

All of the companion's language, and everything written for you, is generated by Anthropic's models. All of the companion's speech, and your meditation and wake-up audio and music, is generated by ElevenLabs. Embeddings, when enabled, are generated by OpenAI. Each of these companies processes your information as our service provider, to provide its features to us. None of them uses your information to train their AI. For our voice provider, ElevenLabs, this covers everything sent to it since October 7, 2026, when we switched off the setting that allowed it; under ElevenLabs' own policy, that switch does not reach conversations it received before then.

We do not use your identifiable conversations, transcripts, notes or voice to train AI models. The App contains no training program and no training opt-in, because there is nothing to opt into. If that ever changes we will tell you first and give you a way to opt out.

11. Who we share your information with

We share personal information only with the providers below, each of which processes it on our behalf to run a part of the App; with Apple and Google as the sellers of your purchases; and, if you choose to link, with the other user, as section 8 describes. No affiliate company of Mazalot receives your information.

Anthropic: every turn of every call with the conversation so far and your briefing; every transcript, for screening, notes and the jobs in section 7, including the Confidant label when you are linked; your chart and the names and charts of people in your list, for readings; the Confidant label of a person linked with you, to write guidance for your astrologer; your first name and note-derived text, for meditation scripts; the scrubbed weekly sample. Purpose: the language model for everything the App writes and says. Anthropic keeps what we send it, and what it sends back, for 30 days and may review it during that time for safety and security; after 30 days it is deleted. Anthropic never uses it to train its models.
ElevenLabs: your live call audio in both directions, which it records and keeps for 30 days, then deletes automatically; the transcript it produces; the call briefing, opening line and voice choice; meditation and wake-up scripts with your first name and note-derived text; the text of your name for the name clip; music prompts; our deletion requests. Purpose: speech recognition, the companion's voice, meditation and wake-up audio, music, and keeping call recordings for 30 days for quality checks and investigating problems.
LiveKit: the live audio stream, as the transport built into the ElevenLabs software.
OpenAI: note text, and during calls some of your spoken words, only when embeddings are switched on. Purpose: embeddings. OpenAI may keep what we send it, and what it sends back, for up to 30 days to check for abuse.
Supabase: all of the data in section 3, your sign-in identity, and stored files (chart image, meditation audio, name clip). Purpose: hosted database, sign-in and file storage.
Vercel: every request the App makes to our servers, including your network address, and the network address of each visitor to the mazalot.ai website. Purpose: hosting our server code, scheduled jobs and the website.
Railway: your birth date, time, place, coordinates and time zone, and your chart for transit calculations; short-lived links to audio for mixing wake-up and event meditations and for adding the name clip. Purpose: our chart calculation and audio mixing service.
Twilio: your phone number. Purpose: sending sign-in codes, and a code that confirms a deletion or privacy request made by email, by text message.
GeoNames: the birthplace text you type while searching, with no account identifier, sent over an unencrypted connection. Purpose: turning a place name into coordinates and a time zone.
Expo, Google Firebase Cloud Messaging and Apple Push Notification service: your push token and each notification's title, text and link. Purpose: delivering push notifications you enabled. Expo also checks for over-the-air updates to the App.
Apple: purchase and subscription events, with your internal account identifier; your public App Store reviews and our replies; the age range Apple reports at your first sign-in. Purpose: billing, reviews, age checks. On iPhone the alarm is set through Apple's alarm system on your device.
Google Play: purchase tokens and product identifiers, with your internal account identifier; your public Google Play reviews and our replies; the age signal Google Play reports at your first sign-in. Purpose: billing, reviews, age checks.
Resend: our staff's email replies to an account email, where one exists; a daily summary to our company inbox quoting low-star reviews and notes, and notes sent while your experience label was poor, without names; staff sign-in codes and alerts. Purpose: email.
Microsoft, our company email service: the emails you send to compliance@mazalot.ai or info@mazalot.ai, with anything you put in them such as your phone number or a store receipt, our replies, and the daily summary described in section 3. Purpose: receiving and answering email.
Sentry: crash and error reports from the App and our servers, when crash reporting is switched on, with user details, request details and activity trails removed before sending. Purpose: finding and fixing crashes.
Google Fonts: your IP address when the website loads its fonts.
WhatsApp or your messaging app: only if you choose to share an invitation that way: the invitation text and the number you send it to, sent from your own phone.

In the categories California law uses, in the past twelve months we have disclosed these categories for business purposes, to the service providers above and only for the purposes named: identifiers; customer records (name and phone number); commercial information (purchases and minutes); internet or other electronic network activity (requests, product events); audio and electronic information (call audio processed live by ElevenLabs, and call recordings, kept by our voice provider for 30 days); inferences (conversation notes, the experience label, Confidant labels); and sensitive personal information, to the extent your conversations reveal it (section 13). We do not collect precise geolocation or biometric information.

We may also disclose information when the law requires it, to respond to a lawful request from a public authority, to protect the safety of a user or the public, to enforce our Terms, or as part of a merger, sale or reorganization of our business, in which case the recipient must honor this policy.

We do not use product analytics or advertising software from any third party, and no third-party software collects analytics or advertising data inside the App. If we add product analytics later, we will update this policy before it starts.

12. No sale, no sharing for advertising, no ads, no tracking

We do not sell your personal information. We do not share it with anyone for cross-context behavioral advertising. We do not show advertising in the App. We do not use advertising identifiers, pixels or tracking software. The mazalot.ai website sets no cookies and uses no analytics; the only outside resource it loads is Google Fonts.

Because we never sell or share personal information, there is nothing for you to opt out of. If your browser or device sends an opt-out preference signal such as Global Privacy Control, we treat you the same as we treat everyone: as already opted out of sale and sharing. Because we do no tracking across sites or apps, a browser's Do Not Track signal changes nothing about how we handle your information.

Giving information to the providers in section 11 so they can run the App for us is not a sale. Linking with another user, which you choose and can end at any time, is not a sale either.

13. Sensitive personal information

Some of what you share is sensitive: the contents of your conversations, what the AI infers about your inner life, and, if you choose to discuss it, your health, beliefs, sexuality or relationships. We use sensitive information only to provide the features you use and to keep you safe, never to infer characteristics about you for advertising. The permission screen in section 6 asks for your agreement before any of your readings or conversations are processed, and it tells you that what you share can include your health and feelings; where a state law requires consent to process sensitive information, that agreement is the consent we rely on. Withdrawing it in Settings stops new readings, calls, meditations and wake-up readings; to have what we already hold deleted, delete your account or ask us (sections 15 and 16). If you live in Washington, Nevada or Connecticut, our Consumer Health Data Privacy Policy explains how we handle information about your health. You can avoid sharing sensitive information by not telling it to the companion; anything you say on a call is transcribed and may appear in your conversation notes, and, if you are linked, may shape the Confidant labels described in section 8.

14. How long we keep it

Section 3 gives the retention for each category. Apart from call recordings, which our voice provider deletes 30 days after each call (section 6), nothing about you is deleted by the passage of time while your account is open, and your account's records are deleted when your account is deleted, with these exceptions:

Product events keep their content but lose your account identifier.
A single record is written at deletion with usage figures (number and length of calls, minutes used, how far setup got, days from sign-up to deletion, plan) and no identifiers.
The deletion log keeps your internal account identifier and the dates of the request and the deletion, so that we can prove the deletion happened.
Emails you sent to compliance@mazalot.ai or info@mazalot.ai, such as a deletion, privacy or support request, a removal request or an arbitration opt-out, our replies, and the daily summaries described in section 3, which stay in our company mailbox (section 11) as our record of what was asked and what we did; the automated deletion does not remove them.
The staff audit log keeps its records of every transcript opening and every listen to a call recording, including your internal account identifier, and our replies to feedback keep their text.
The de-identified crisis referral records in section 3.
Store reviews you posted publicly, with our guess at your account removed.
Entries that other users made about you in their lists, and invitations others sent to your number, as section 8 explains.
The abuse-prevention table holds phone numbers and network addresses for about two days after each sign-in attempt, sometimes longer before cleanup.
The place-lookup cache keeps birthplace search text, with no account identifier, without a time limit.
Copies outside our live systems for a limited time: our providers' backups and logs, which they overwrite or delete on their own schedules, and what an AI provider keeps under its own terms for abuse monitoring, such as OpenAI's copies for up to 30 days (section 11).
Records we must keep to comply with the law, for as long as the law requires.

15. Deleting your account, and getting a copy of your data

Deleting in the App. Open Settings in the App, choose Delete account, then confirm. You do not need to contact us. The App signs you out and clears the notification token on your device.

Deleting without the App. Follow the steps at mazalot.ai/delete-account. Email compliance@mazalot.ai from any address with the phone number you sign in with. Before we act, we send a one-time code by text message to that number, and you send the code back to us, so that we know the request comes from the person who holds the number. We then start the deletion exactly as if you had asked in the App.

What happens next. The account enters a thirty-day grace period, during which it stays usable. If you change your mind, write to compliance@mazalot.ai within the thirty days and we will cancel the deletion. Once the thirty days pass, an automated job deletes the account and its records: your identity and phone number, birth data and chart, transcripts, conversation notes and everything generated from them, the people in your list and the notes about people you mentioned, your links and Confidant records, meditations, purchases and the minutes ledger, feedback, notification settings, and your sign-in identity. It also asks ElevenLabs to delete each of your conversations, including any call recording still kept, right away; if that request fails, we retry every hour until ElevenLabs confirms it is done. Your chart image, meditation audio files and the clip of the companion saying your name are removed from our private file storage as part of the same deletion, at the latest seven days after the automated job runs; you do not need to ask. The exceptions in section 14 are the only records that remain. Deletion is permanent. Apple and Google Play keep their own records of your purchases.

Deleting your account does not cancel your subscription; cancel it in your Apple or Google Play subscription settings first.

A copy of your data. The App does not have a button that exports your data. To receive a copy of the personal information we hold about you, email compliance@mazalot.ai with the phone number on your account. We will verify the request and provide the copy in a readable, portable format within the time limit in section 16.

16. Your privacy rights and how to use them

Whatever state you live in, we give you these rights:

To know what personal information we have collected about you, where it came from, why we use it and who we share it with, and to receive a copy of it.
To correct inaccurate personal information. Your name is corrected by signing in with the right name; anything else, ask us.
To delete your personal information, subject to the exceptions in section 14.
To take your data with you in a portable format.
To limit the use of your sensitive personal information to what is needed to provide the App, which is already all we do.
To withdraw an agreement you gave: the permission screen in Settings (section 6), and a link by unlinking (section 8).
To appeal if we decline a request, and to complain to your state attorney general if you disagree with our decision on the appeal.
Not to be treated differently for exercising any of these rights.

To use a right, email compliance@mazalot.ai or write to the address in section 1. Tell us which right you are using and the phone number on your account. We will verify that the request comes from the account holder, usually by sending a one-time code to that phone number. An authorized agent may make a request for you if they provide your signed permission and we can verify your identity. We will respond within forty-five days of receiving a verifiable request. If we need more time we will tell you within those forty-five days and may take up to forty-five more. To appeal a decision, reply to our answer with the word "appeal" and your reasons; we will answer the appeal in writing within forty-five days of receiving it.

California residents may also request, once a year, a list of the third parties to whom we disclosed personal information for their own direct marketing purposes in the previous year. The answer is that there are none; we make no such disclosures.

17. Notifications, text messages and the horoscope alarm clock

Text messages. The only text messages we send are one-time codes: the sign-in codes, and a code to confirm a deletion or privacy request made by email. We do not send marketing texts. Carrier message and data rates may apply. The Terms explain how codes work and how to stop them.

Push notifications. If you allow notifications, the App stores a token that lets us send messages to your device through Expo, Google Firebase Cloud Messaging and Apple's push service. The App can send a daily reading notification whose title and text come from the opening of your reading, a short check-in ring at times you scheduled, a note about a meditation made for an event you mentioned, and replies from our staff to feedback you sent. You can turn the daily reading notification off in Settings, and every notification off in your phone's settings. Notifications that cannot be delivered within twenty-four hours are discarded unsent. When a gift program is on (Terms section 11), the App may also send a notification about a gift, unless you have turned notifications off.

Horoscope alarm clock. On Android phones and on iPhones running iOS 26 or newer, if you turn on the wake-up alarm, our server prepares the audio the night before and your phone stores a private link to it. The alarm is set on your phone, by your phone's own alarm system, at the time you chose; on iPhone the App first asks your permission to set alarms. When it rings, your phone plays the reading out loud; on Android it can appear over the lock screen and the alarm is restored after a restart. We do not trigger the alarm remotely. Turning the alarm off cancels it. On older iPhones the alarm is not offered.

18. Crash reporting and product analytics

Crash reporting. The App and our servers are set up to send crash and error reports to Sentry. A report contains technical details of the failure, the app version and platform; before it is sent, user details, request details and activity trails are removed.

Product analytics. We record product events in our own database, as section 3 describes, and they do not leave our systems. We have not connected a third-party analytics service. Our rules for product events forbid transcript text, names, phone numbers and email addresses in them. A few events carry short generated labels about a call, for example the chart placement the companion referred to, and one event carries a short error message when after-call processing fails.

19. What happens to crisis information

Our server checks every turn of every call, using fixed rules, for signs of self-harm or suicidal thinking. When it detects them, the companion sets aside its normal reply and speaks a fixed message that names 988, and the App shows a crisis card with buttons to call or text 988. The words of the call are kept in the transcript like any other call. If after-call processing finds signs of crisis, the call's notes carry a crisis flag that stops them being reused in cards, the Confidant and secondary features, and a product event records that flag with your account identifier until your account is deleted. If your next daily reading comes within 72 hours of such a call, it opens with a fixed message and 988 instead of a normal reading. We also write a de-identified referral record with the date, the kind of statement matched and a two-letter region, and nothing else, and we keep those records to prepare the annual reports that California and other states require. The full protocol is published at mazalot.ai/crisis. We never use crisis information for marketing or to make decisions about your account.

20. Security

We protect your information with access controls, passwordless sign-in, rate limits and lockouts on sign-in attempts, signature checks on the messages our voice provider sends us, private storage for your files reached only through short-lived links, encrypted connections for our app and server traffic, a restricted database role that is the only way a Confidant label can be read, and the staff access controls and audit log in section 9. One provider, GeoNames, is reached over an unencrypted connection, and only the place text you type is sent to it. Our staff dashboard requires a one-time code. No system is perfectly secure and we cannot promise that yours will never be breached. If a breach affects you we will notify you as the law requires.

21. Children

Mazalot is for adults. You must be 18 or older to hold an account. The App checks the birth date you enter for your own chart and refuses anyone under 18, and at your first sign-in it also reads the age range your app store reports, if any, and refuses an account the store reports as under 18. We do not knowingly collect personal information from anyone under 18. If we learn that an account holder is under 18 we will delete the account. If we learn that we have collected personal information from a child under 13, we will delete it immediately, as the Children's Online Privacy Protection Act requires. If you believe a minor has an account, write to compliance@mazalot.ai.

22. Where your information is processed

Mazalot operates from the United States and your information is processed there and by the providers in section 11. The App is offered in the United States only. If you use it from another country, you understand that your information is transferred to and processed in the United States, where privacy laws may differ from those of your country.

23. Changes to this policy

When we change this policy we will post the new version at mazalot.ai/privacy with a new date at the top. For a material change, we will tell you in the App before the change takes effect. If a change would let us use information we already hold about you in a materially new way, we will ask for your consent where the law requires it. A change to what linking shares does not apply to a link until both people agree to it (section 8.6).

24. Contact

Mazalot Inc., 11631 Victory Blvd Ste 201, North Hollywood, CA 91606, United States. Email: compliance@mazalot.ai. Telephone: +1 (323) 330-5122.