Privacy Policy
Effective date: November 16, 2026 Last updated: November 16, 2026
1. Who we are and what this covers
Maison Mason LLC, a New York limited liability company doing business as AroundBack ("AroundBack", "we", "us"), operates the AroundBack iOS app and the AroundBack service. This policy covers the app, our servers, and our web pages at aroundbackapp.com. Our moderation and admin consoles are internal staff tools and are covered here only insofar as they display member data to our own moderators.
AroundBack is a location-aware social and dating app for LGBTQ+ adults. That fact shapes everything below: for many of our members, the existence of an account is itself sensitive, and in some places it is dangerous. The design decisions described in this policy (coarse location, end-to-end encryption, on-device journaling, no analytics vendor) were made for that reason and not as marketing.
AroundBack is offered in the United States and Canada and is for adults 18 and older.
Consumer health data is covered by a separate policy. See our Consumer Health Data Privacy Policy, which has its own link on our homepage.
Contact for anything in this policy: privacy@aroundbackapp.com · Maison Mason LLC d/b/a AroundBack, Legal Affairs, 2248 Broadway #1231, New York, NY 10024.
2. The short version
- We hold no readable copy of your messages, calls, private photos or private videos. They are end-to-end encrypted; our server relays ciphertext it cannot open.
- We hold your location as a coordinate only briefly and only in memory. Nothing we write to disk can locate you: your coordinate is coarsened to a roughly 1 km grid cell center first, or left out of the file altogether. (The file does hold exact coordinates for one thing: the published street addresses of the bars, clinics and community spaces in our venue directory. Those are facts about businesses, not about you. Some of the things you do in the app link you to one of those addresses; a record of that kind is kept on a published clock and then deleted, rather than indefinitely. Section 3.3a lists which records those are and how long each is kept, and that list is generated from our code rather than typed out here.) In discovery, another member never receives a coordinate or a distance in meters, only a band label like "Nearby" or "1–3 mi". A Now you post can also appear on the Now map as a dot at your roughly 1 km cell; Section 3.3 explains. When you deliberately share your live location with someone (a Guardian check-in or an On My Way trip) they do see where you are, because that is the point, and it is deleted when the share ends.
- Your Encounter Journal stays on your phone unless you choose to back it up, and even then we receive only a blob we cannot decrypt.
- We run no analytics SDK, no advertising SDK, and no ad system. The server has zero third-party runtime dependencies. Our crash and performance reporting is our own, opt-in, and anonymous at ingest.
- We do not sell your personal data and we do not share it for cross-context behavioral advertising.
- You can export the data we hold about you and delete your account from inside the app, at any time, on any tier, for free. The export leaves out reports other members filed about you, which are our records rather than yours, and gives a few records as a summary rather than in full; Section 9 says which.
Each of these is stated more precisely, with its limits, below.
3. What we collect
This section is our notice at collection: the categories of personal information we collect, where each comes from, and what we use it for. Everything we collect comes from you directly, from your device with your permission, from Apple in the course of sign-in and purchases, or, when another member reports you, from that member (Section 3.5). We buy nothing about you from anyone, and we collect nothing about you from data brokers, social networks, or public records. There is exactly one exception, and it is not about members: when a business sends us a sponsorship enquiry for a charity event, a member of our team may look the beneficiary up in public records, as Section 3.9 describes.
3.1 Account and identity
- Account identifier. An opaque id (`acct_…`) we mint for your account. For Sign in with Apple it is a one-way hash of Apple's subject identifier for you, so signing in again with the same Apple ID returns the same account — deliberate, so a removed member can't shed their history by re-registering. We store that subject identifier too (see the next item). It is a stable pseudonym, not an anonymous token.
- Sign in with Apple. If you sign in with Apple, we store Apple's stable subject identifier for you, and the email address Apple releases (which may be Apple's private relay address). We verify Apple's signed identity token against Apple's published keys. We also keep the refresh token Apple issues to us at sign-in, encrypted, for one purpose only: when you delete your account, we use it to revoke your Sign in with Apple authorization at Apple, as Apple requires, and then discard it. We never use it to read anything from Apple.
- Passkeys. If you register a passkey, we store its credential id, its public key, and a signature counter. We never receive a private key. Face ID and Touch ID unlock the app locally on your device and are never used to identify you to us.
- No passwords. We do not offer password sign-in, so there is no password of yours for us to hold or leak.
- Sessions. A hash of your session token, its creation and expiry time, and its scopes. Once a session registers the phone it runs on as one of your devices (Section 3.7), we also record that device's id on the session, so that removing the device from your account signs that phone out.
- Age. Whether you are 18 or over. The app first asks Apple's Declared Age Range, on your device. If Apple gives a range, that range decides: 18 or over passes, and under 18 ends sign-up before any account exists (we keep only a daily count, with nothing about you). If Apple gives no range, you confirm that you are 18 or over, and the live check's on-device estimate clears you when it confidently reads you as an adult; otherwise, after a second look, a person on our team reviews your profile photos (and, only if you agree, one photo from the check; see Verification and the live check). If that review cannot confirm that you are an adult, the account is closed and its data deleted, and sharing an Apple age range is the way to ask again. Optionally, a coarse age band (for example "25–34") if you complete the age range flow, and a numeric age if you enter one yourself (you choose who can see it). Apple's signal is a coarse range, never a birth date. If an age block reaches your account after sign-up, you can appeal to a person, and we record why the block happened. We do not collect your date of birth, we never ask for an identity document, and we do not claim to have verified every member's age; see Section 11.
- Recovery. If you enable them: one-time recovery codes (stored hashed) and, if you enable social recovery, the account ids of the trusted contacts you nominate who are members, a hash of the phone number or email you give for a contact who is not (rather than the number or address itself), and blinded shares that we cannot reassemble alone.
- Consent records. Each answer you give to a consent question, with the policy version and the time, tied to your account id.
3.2 Profile
Everything in your profile is written by you and is optional beyond a display name: headline, bio, a coarse self-chosen neighborhood, pronouns, intents, tags, body type, position, height, relationship status, ethnicity, prompt answers, your stated boundaries, desires and intentions, birthday (month and day only, never a year), photos, and a voice intro. You control per-field visibility (everyone or only me), and Anonymous mode withholds your photos, name and free text from viewers entirely.
Because AroundBack is a dating app for LGBTQ+ adults, information you put in your profile (and your membership itself) can reveal or imply sexual orientation, gender identity, racial or ethnic origin, and related characteristics that state privacy laws treat as sensitive personal information. We ask for your consent before we process the optional profile fields that carry this kind of information, we use them only to run the service you asked for, and you can withdraw that consent in You, App & data, Privacy choices at any time (withdrawing removes the data the consent covered). We never use these fields for advertising, and we never sell them; see Section 7.
Note honestly: prompt answers, tags, your bio and your stated boundaries, desires and intentions are free text. Whatever you type there is readable by us and by whoever you make it visible to. If you write health information into a prompt, it becomes health information we hold; see the Consumer Health Data Privacy Policy.
3.3 Location
This is the part of the product with the most engineering behind it, so it is worth stating exactly. Precise location is sensitive personal information under many state laws. We ask for your consent in the app before discovery uses it (in addition to iOS's own location permission), and you can withdraw that consent, or turn location off in iOS Settings, at any time.
What your phone sends. When you use discovery, the app posts your coordinate to us. The response contains no echo of it. Location is captured only while you are using the app, at reduced accuracy (roughly 100 m), and your device applies your Privacy Zones cloaking before anything leaves it. The only features that use background location are the ones where you deliberately share a live trip (Guardian check-ins and On My Way), described below.
The emergency number. Safety & Support, and the Safety screen on Apple Watch, use the location of your phone or watch to show the emergency number where you are, and to avoid suggesting the police where that could put you at risk. This happens on your device only: the app looks up which country you are in using a map and a list of numbers that are stored inside the app. The app asks iOS for your location for this only when you tap the control that says so (on Apple Watch, when you first open the app's glance screen and when you open its Safety screen), and it also uses the most recent position your phone already has, including one the app received for discovery. The position is held in memory only, is not saved to a file, and is gone when the app closes. We do not receive it, and nothing about it is sent to us or to anyone else. This use runs on iOS's location permission and does not depend on the precise-location consent you give us for discovery: withdrawing that consent does not turn it off, and turning off location for AroundBack in iOS Settings does. When the app does not know where you are, it shows 911 and says so.
What we do with it.
- We hold the raw coordinate in server memory only, keyed to your account, with the time of the fix and the accuracy your phone reported.
- Before any other member sees anything derived from it in discovery, we snap it to a grid cell (roughly 1 km across), take the cell center, add a deterministic offset of up to a few hundred meters that is unique to the viewer, and turn the result into a band label. The viewer receives the label. In discovery they never receive a coordinate, and never a distance in meters. (The deliberate live shares described below are the opposite by design: there, the person you chose does receive your position.)
- In dense areas the grid can tighten to roughly a half or a third of that, which is a deliberate trade for usefulness in cities. That finer resolution is only granted to a viewer who has stayed in their own reported cell for 15 minutes, is capped at two distinct such vantage points per hour, and requires a verified viewer, because three vantage points is what trilateration needs. This raises the cost of an attack; it is not a proof of impossibility, and we say so.
- Nothing we write to disk can locate you. Before the store is serialized, every member location in it is replaced by its cell center. This is band-preserving, so it changes nothing you see. It holds for the whole file, not just discovery: the deliberate exceptions below are exact in memory and are simply left out of what we write down. The file does contain precise coordinates for one category of thing: the published street addresses of the venues in our public directory, which are facts about businesses, not fixes taken from anyone's phone. We say so rather than claiming "no precise coordinates anywhere", because a claim you can disprove by opening the file is worth less than a narrower one that is true. We check this with a test that serializes the real database and looks for coordinates by the shape of their numbers rather than by what a field is called. It is a best-effort sweep, not a proof: Section 3.3a below lists the shapes it recognizes and the shapes we know it misses, because a check described more confidently than it works is worth nothing.
- And we will not leave the rest of that sentence for you to work out. Those venue addresses are exact and public, and some of the things you do in the app link you to one of them. An intention to be at a named address at a named time is, for many of our members, more sensitive than a coordinate would be, and one of these is not an intention at all, but you telling us you are there right now. None of the records we know of is kept indefinitely. That rule (a record we keep that links you to a venue is kept on a bounded clock) is the commitment, and it is meant to bind the records we add later as much as the ones we have. The list of the ones we have today, with each clock, is in Section 3.3a below; it is generated from our code rather than written out by hand.
- A background sweep coarsens any coordinate still in memory to its cell center once the fix is more than 6 hours old. The same clock and the same sweep apply to a Travel Mode pin you drop by hand; it is a discovery origin like any other.
The deliberate exceptions are all of them you sharing your own location on purpose; the ones that exist today are listed below. In every case the coordinate is exact, is held in memory only, is never written to disk, and is deleted (not merely hidden) when the share ends. That rule, rather than the length of this list, is the commitment: any feature we add that lets you share your live location inherits it.
- Guardian meetup check-ins. You choose a trusted contact and share your exact, live location with them until you mark yourself safe. That contact is never told who you are meeting. If they are not an AroundBack member, we store no contact details for them at all; you hand them a private status link yourself. If you start a check-in with location sharing switched off, we store no coordinate for it at all. The live fix is never written to disk, so if our server restarts your contact sees "no location" until your phone sends the next one, rather than a stale or approximated pin. When you mark yourself safe, cancel, or switch live sharing off mid-trip, we delete the coordinate at that moment. The check-in record itself stays visible to your contact for a while afterwards so they can see the trip closed cleanly, but it no longer holds a position. The one deliberate exception is the duress code: that keeps your location flowing to your contact, because the whole point is that you may need to be found. Once the trip is over, the link shows only that it is over: your name and your meeting note come off it. The link itself carries its secret in the part of the address a browser never sends to a server, so it does not land in web server logs or in the previews messaging apps fetch; you can reset it mid-trip if you think it was forwarded, which stops the old one at once; and you can set a PIN on it that the page asks for once in each browser, which you tell your contact separately. So that you can see whether the link was forwarded, the page sets a random cookie in your contact's browser that identifies nothing but that browser, and we keep, for you alone, when the link was first opened and how many different browsers have opened it. The cookie value and the first-open time are the only things we hold about your contact's browser; we store no address and nothing else about the device, and they are deleted with the check-in. The page says this to your contact. If a second browser opens it, the app tells you. The page shows your last position as a coordinate with two buttons, Open in Apple Maps and Open in Google Maps; the page itself loads nothing from either company, but if your contact taps one, that company receives the coordinate. If your contact opens the link on an iPhone or iPad, an App Clip (a small piece of the app that needs no install and no account) can show the same page and, only if they tap Notify me in it, alert their phone if you do not check in or you signal that you need help. While the clip is open and your check-in runs, it can also show the check-in on their Lock Screen, which shows the time you are due and never your name or a place. For that we hold, for their phone, a notification token and, while that Lock Screen display runs, a token for the display, both tied to that one check-in and deleted with it, when you reset the link, or when they turn it off in the clip; a notification token that rests only on the App Clip's temporary permission is deleted eight hours after it was registered. The clip collects nothing else about them, and the app shows you whether a phone will be alerted, never which. If your contact is an AroundBack member, they can show the check-in on their own Lock Screen from the app in the same way; the token for that display is tied to the check-in and deleted with it, if either of you blocks the other, if they delete their account, or if they sign out of the phone that showed it.
- On My Way. Started from a chat with the person you are meeting, this shares your live position with them until you arrive or stop it. It is the one place in the app where another member is given real distance and direction: far away they see a rounded distance and a compass direction, closer in they see an approximate pin, and within about 200 meters they see your exact position. That is what the feature is for; it is one-way, you start it, and either of you can stop it. It is capped at two hours, it stops automatically once your phones have been together for ten minutes, it is severed immediately if either of you blocks the other, and your position is deleted when the trip ends. Nothing from it is ever written to disk.
The Now map never shows where anyone is standing. Where five or more people are in a grid cell, it shows aggregate "energy": counts are buckets ("a few", "busy") rather than numbers, and the bucket edges are blurred for each viewer. A quieter cell is merged into a larger area, and where there are still too few people, those who are simply online are not shown at all. A Now you post is different: in a cell with fewer than five people, a Now posted with Show on the map switched on (it is on unless you switch it off when you post) appears as its own dot. The dot sits at the center of your roughly 1 km cell, moved by an offset that is different for each viewer, so the finest anyone can learn from it is that cell, never your position. Other members can tap the dot to open your Now, which shows who posted it unless it is a discreet Now. A Now posted from inside one of your Privacy Zones, or with Show on the map switched off, stays off the map. A venue shows a count of the Nows anchored to it only once there are at least three.
3.3a What links you to a venue, and for how long
A list written out by hand in a document stops matching the software the moment anyone adds something.
So this list is not written out by hand. It is generated from one table in our code, the same table our deletion sweeps read their clocks from, and a test fails our build if this page and that table disagree. That much is mechanical, and it is worth being precise about what it does and does not buy you:
- What it covers. The windows in the rows below are written from the same constants our deletion sweeps read, so they are the numbers the software uses rather than a copy somebody typed here, and a change to that table (a new row, a shorter or longer window) turns this page red until it is regenerated.
- What else we do. A second test reads our real stored database file and looks for any saved record that carries both an account id and a venue or event id. If it finds one the table does not name, the build fails. That is the check meant to catch a new kind of link we forgot to write down here.
- What that second check does not reach. It only sees records that are actually written to the stored file, in the test's own run, with the ids recognizable as plain text. A link we hold only in memory, one the test's scenario never causes to be written, one where the account or venue id is hashed or packed into some other format, or one recorded somewhere other than that file, would not be found by it. We would rather say that than let "a new kind of link cannot land silently" stand, because it is not something a test can promise.
| What links you to a venue | What it says | What kind of statement it is | How long we keep it |
|---|---|---|---|
| An event RSVP | the event, and so the venue it is held at | An intention (we never learn whether you went) | Deleted 7 days after the event ends |
| A saved event | the event you bookmarked, and so the venue it is held at | An intention (we never learn whether you went) | Deleted 7 days after the event ends |
| A Now drop anchored to a venue | that you said you were AT that address, at that moment, next to whatever note and intent you posted — and it says so even for a discreet drop, where the mask other members see is applied when the drop is read, not when it is written | You saying you are there (not an inference — you posted it) | Deleted when the drop expires — at most 12 hours after you post it — by a background sweep, so it goes whether or not anyone opens the map |
| A listing report you filed | the venue or event you reported, and so its address, and any words you added to the report; nothing about whether you were there. The report itself is kept for the listing's sake; what is deleted is the link to you and the words you wrote | A report you filed (it says nothing about whether you went) | Your id is removed from it, and the words you wrote are deleted, 30 days after the report is resolved, and at the latest 90 days after you filed it whether or not it was resolved |
Deleted means deleted: a background sweep removes the record from the store, so it leaves the database file too, rather than being hidden from a screen. Deleting your account removes the records in this table at once. The one exception is a listing report you filed: the report is kept for the listing's sake, and it is your id that is removed from it, at once on deletion and on the clock above otherwise.
This table is generated from one table in our source code rather than retyped here, so the rows above and the clocks our deletion sweeps actually run on are the same text. What that does not by itself prove is that the table names every such record — the separate check for that is a scan of the stored database file, and its limits are described in §3.3a of the privacy policy.
How we check the coordinate promise, and what the check misses. The test described in Section 3.3 point 4 serializes the real database and looks for coordinates by their shape rather than by what a field is called: where two numbers sit side by side on the same record, land in valid latitude/longitude range, and carry street-level precision, it treats them as a coordinate and requires them to be a grid-cell center, whoever wrote them and whatever they are named. That closes the obvious way around a check like this, which is to call the fields something new.
It is a best-effort sweep and not a proof, and we would rather list its blind spots than let you infer it has none. We know of stored shapes the sweep does not flag, and each is written into our test suite as a known limit, so they are recorded in code rather than only in prose:
- a position stored as whole numbers in a scaled unit (millionths of a degree, a packed geohash integer);
- a position stored as a list of three or more numbers, such as a latitude, a longitude and an accuracy figure, rather than a plain pair;
- the two halves wrapped in separate sub-objects, so no single record has two loose numbers side by side;
- both halves inside one piece of text, such as `"40.712834,-74.006177"`;
- an offset or otherwise transformed pair that lands outside normal latitude/longitude range but can be transformed back;
- the two halves stored in separate records and joined by a key, or hidden inside data that is encrypted or otherwise encoded.
We could widen the rule to catch some of these, and chose not to: each widening also flags ordinary numbers (byte counts, offsets, identifiers, measurements), and a check that cries wolf gets an exception list, which is precisely how a real leak would one day get waved through. So the honest scope is narrower than "we would detect any coordinate". What this check does cover is a coordinate written as a plain pair of numbers, under any field names, anywhere in the file, including next to your account id, which is the shape careless code actually produces.
Photos and voice intros cannot carry embedded location off your device. The two media types are handled differently, and the difference matters:
- Photos are re-encoded, not inspected. Our server re-muxes every uploaded JPEG through an allowlist, keeping only the image data itself and physically dropping every metadata segment, which is where Exif, XMP and GPS live. We do not scan for GPS and reject it; a list of known-bad shapes is something crafted files slip past. Removing the container the location could ride in is the stronger guarantee, and it is the one we make. A file we cannot parse as a JPEG is refused.
- Voice intros are inspected and refused. An audio file carrying a location atom is rejected outright, because an audio container can be re-muxed without changing what the recording is, so we can afford to be strict.
In both cases we do not trust the app's own stripping: the server does its own.
3.4 Messages, calls and media
- Direct and group messages are end-to-end encrypted. Our server stores ciphertext and opaque routing headers. It holds no message keys. Our main message route refuses any send that carries a plaintext-shaped field; the sealed and discreet message routes read only the ciphertext and the routing fields they need, and any other field a request carries is dropped unread, never stored and never delivered.
- What we necessarily do see for messaging to work: that an account has a conversation with another account, message sizes, and timing. This is metadata, and it is the honest residual of any relay. We minimize it (dead conversations are forgotten entirely after the period in Section 8, and sealed sending hides the sender from us on the paths that use it), but we do not claim to hold none.
- Attachments (photos, video, voice notes in chat) are uploaded already encrypted by your device. We store opaque bytes we cannot decrypt, plus a reference, an expiry, the size and length, whether it is a view-once item, a flag telling the recipient's app to blur it, and, for a photo or video, the fingerprint described in Section 3.5, which we keep so a later takedown request can find the copy.
- Private Media (your private album) is encrypted to a specific recipient. We hold ciphertext and consent state (who asked, who granted, when it expires, whether it was revoked), whether your phone flagged the photo as sensitive (Section 3.5), and the fingerprint described in Section 3.5.
- Signals. When you send another member a Signal (such as a Hey or a Date), we keep a record of it: an id for the record, which kind of Signal it was, the account ids of the sender and the recipient, the time it was sent, and when its 8-second undo window closes. A Signal you undo within that window is deleted. Otherwise the record has no time limit: we keep it until either of you deletes your account. If either of you blocks the other, the record is not deleted: we stop listing it to either of you, and it appears only in your own data export (Section 9.1).
- Calls. We relay signaling metadata (who is calling whom, when, and the call's state) and mint short-lived credentials for the media relay. We never parse the session description, and the media itself is relayed as encrypted RTP. We do not record calls, and the product has no way to record one.
- Public profile photos and voice intros are the one place we hold plaintext media, and we hold it because a public photo has no single recipient to encrypt it to. Because those bytes are readable by us, screening and access gating on them are mandatory rather than optional.
3.5 Safety, moderation and anti-abuse
- Reports you file: who and what you reported, the reason, and any free text you write. Free text in a report is readable by our moderators. A report is also a record about the member it names: it sits on a moderation case about them, and one reason on the list, "Worried about them", is a statement about that member's mental health made by someone else. The Consumer Health Data Privacy Policy, Section 2.3a, describes that reason, what it can and cannot cause, and how it is severed from a deleted account.
- Message reports carry a cryptographic commitment that lets a moderator verify a reported message was genuinely sent, without us holding the conversation.
- Blocks, mutes, hides, pauses, and your visibility settings.
- Moderation records: cases, decisions, appeals, and an append-only audit log of moderator actions.
- Anti-abuse signals: account age, count of reports received (accusations only; a "Worried about them" report is not counted), prior enforcement, flood events, suspected location spoofing, and links between a device and the accounts on it (this is our defense against one person running many accounts).
- Device attestation. We verify Apple's App Attest assertion for your app instance. It proves the app is genuine on a real device; it is not a device fingerprint we can use to recognize you elsewhere.
- Image hashes for known-CSAM and nonconsensual-imagery screening. A hash, not the image. Two different mechanisms, and we will not blur them:
- CSAM screening of public profile photos and voice intros uses exact cryptographic digests (MD5, SHA-1, SHA-256) that our server computes from the bytes you upload and compares against material our moderators have already confirmed; no outside hash list is loaded yet. Exact digests catch an unmodified re-upload and are defeated by a re-encode or a crop. We do not run a perceptual (PhotoDNA-class) matcher; that is a licensed service we have not deployed.
- Nonconsensual-imagery matching uses a perceptual hash your device computes before the photo is sent, checked against a takedown registry.
- On end-to-end-encrypted media (chat attachments and private album photos and videos), the only check is a fingerprint your device computes. Before encrypting a photo or video, the app computes a perceptual fingerprint on-device and sends the fingerprint (never the image) with the upload. We match it against the nonconsensual-imagery takedown registry; a match is refused and the stored bytes are destroyed, and a verified match is treated as actual knowledge and handled as the law requires. We do not screen encrypted media against the verified child-exploitation hash lists, and with the fingerprint alone we could not: those lists hold exact cryptographic digests of the file itself, we never hold the file, and the fingerprint your device sends is a different kind of value that cannot be compared against them. We cannot see the image itself and cannot verify the fingerprint against it, so a modified app could omit or falsify it: this check raises the cost of abuse rather than proving its absence. The other controls on encrypted media are the on-device screen before encryption, report-driven human review afterwards, and the quarantine of anything confirmed.
- Public free text (such as Now notes) passes an automated screen for threats, scams, hate and attempts to move people off-app for fraud, and may be reviewed by our moderators. We do not decrypt or read end-to-end-encrypted messages or media, and nothing on our servers scans them; the fingerprint check above is the only server check that touches encrypted media. Some safety checks run only on your iPhone. In a one-to-one chat, before you send a message, the app checks your text for threats to share intimate images, abusive language, and anything that crosses a limit the other person has written on their profile, and asks you to confirm before it goes. In a one-to-one chat, when AI features are on (You, App & data, AI features) and your iPhone supports Apple Intelligence, the app also checks the latest message you receive for scam and sextortion patterns and warns you if it finds one. If you have turned on Apple's Sensitive Content Warning in iOS Settings, your phone checks a photo for nudity before you send it in a chat or add it to Private Media, and asks you to confirm before sending one it flags, and it checks photos you receive so it can blur them until you tap. None of these checks stops you sending or viewing anything, and none of them sends us any message text. The one result that reaches us is for a photo you add to Private Media: whether your phone flagged it is sent with it, so the viewer's blur setting can apply.
- CSAM. If content matches a verified hash set or a moderator confirms it, we are legally required to report it to the National Center for Missing & Exploited Children (NCMEC) and to preserve the associated content and metadata in isolation, as described below. That report carries what we hold about the account at that moment: the account identifier, the email on the account, the display name, the network address the upload came from and when, the country our network provider associates with that address, and the approximate location we hold for the account (the roughly 1 km cell, never a precise position). It never carries anyone else's data. We keep that preservation in isolation for one year after NCMEC acknowledges the report, and longer only while legal process or a legal hold requires it; it is not subject to your deletion request. We also keep, in the same isolation and for the same period, a separate internal copy of the account data we held when the report was made. When that period ends we destroy the preserved content and that copy and remove the identifying details from our record of the report, keeping only the file's digital fingerprint, the report number and the dates, so the same material cannot be posted again. Your use of AroundBack is subject to the screening described in this Section 3.5; the Terms of Service contain your consent to it.
- Takedown requests. If you ask us to remove an intimate image, in the app or through our web form, we keep the details that identify you (your signature, which in practice is your name; how to reach you; your statement; and where you saw the image) for 90 days after we have dealt with your request, long enough to answer you and to cover a dispute about what we did, or less if you delete your account. We keep, with no end date, a record that the request arrived, when, what we did and the image's fingerprint; it names nobody, and it is what stops the image being sent again.
- The Hype. When you and another member both confirm that you met in person, we record that the two of you met. Your phones can also confirm it for you: while you are messaging each other or on an On My Way trip together, each phone checks whether the two of you have been within about 300 meters of each other for more than 15 minutes, and if so tells us you met, without sending us any location. (To make that check, each phone adds its position, rounded to about 100 meters, to the end-to-end-encrypted messages it is already sending the other person; we cannot read it.) For 7 days after that, either of you can give the other up to three fixed endorsements (as pictured, respectful, felt safe). We keep the record of the meeting for 8 days and then keep only a count of your own meetings that names nobody. We keep an endorsement until it is withdrawn (a withdrawn one for 30 more days) or either of you deletes your account. Other members see only totals, and only if you switch the display on and at least five different people have endorsed you.
Verification and the live check
What it is. The live check is a short camera check that runs on your device. It confirms that a live person is holding the device, compares your face with the profile photos on your account, and estimates your age. Every new account is asked to complete it once, as the last step of sign-up; what counts is completing it, not passing it. You can also run it from You, then Account and security, then Verification, and we may ask you to run it again, for example after you sign in on a new device.
What stays on your device. The camera frames, depth readings and face measurements the check uses stay in your device's memory and are discarded when it ends. We never receive a face template, a face map, a depth scan or a video of you, and we never compare your face with anyone else's.
What we receive. After each check, a result your device signs so that we can tell it came from the AroundBack app on your device: whether it passed and, if not, why; which of your profile photos matched; your estimated age in whole years, how confident the estimate was, and whether it read as clearly over or clearly under 18; Apple's age range and how it was set, if you shared it; details about how the check ran, such as your device model and operating system, security readings about the device and the app, image quality readings and timings; and the version of the consent you agreed to, and when. We keep it with your account as a safety signal (Section 8).
A check held for review. If the check cannot confirm your age and you have not shared an Apple age range, a person on our team reviews your profile photos. At that moment the app may also ask whether you want to share one photo from the check and the check's detection results with that person. It shows you the photo first. If you agree, your device seals them so that only our review team can open them; our servers store them sealed and cannot open them; the team uses them only for that review; and we delete them no later than 30 days after they arrive. If you choose Don't share, nothing from the check is sent, and choosing it is never held against you. We never ask for this photo if the check estimated that you are under 18.
What we never do. We never ask for an identity document, and we never hold a video or phone call to verify you. We never use the check to sign you in, for advertising, or to build a profile of you, and we never sell anything it produces. Verification is never paywalled, on any tier.
The badge. A passed check shows a verified badge on your profile. Signing in on a new device, or not running the check again by the date we ask, pauses the badge until you pass again; a pause affects only the badge. If you change your lead photo after a pass, other members see that it is not yet matched until you match it.
Our biometric data policy (aroundbackapp.com/biometric-policy) explains this in full, with its retention and destruction schedule.
3.6 Purchases
If you subscribe or buy a one-off item, Apple processes the payment; we never see your card details. We receive and verify Apple's signed transaction and App Store server notifications, and we store what tier you are entitled to and until when. Each purchase carries an app account token — an opaque identifier we derive one-way from your account id and hand to Apple with the transaction. It tells Apple nothing about you; it exists so that Apple's later notices about your account (a refund, or a parent or guardian withdrawing consent under a state age law) can be matched back to it and honoured, even when the purchase record itself is not enough. We do not currently offer purchases outside the App Store; if we ever add a web checkout with a payment processor, this policy will be updated first.
3.7 Device, diagnostics and network
- Push tokens for your devices, so we can wake the app. If you show a Guardian check-in you look out for on your Lock Screen, we also hold a token for that display, tied to that check-in and deleted with it (Section 3.3). Push payloads carry no message content, only a hint that something arrived. The one exception is an incoming call while the app is not open: that push carries the caller's display name (or "AroundBack call" if the caller uses Anonymous mode) and account id, so your phone can show who is calling.
- Devices. For each phone signed in to your account: an id for the device, its model (for example "iPhone"), when it was added, and its public encryption key with a fingerprint of it, so you can see and remove your devices.
- Crash, hang and app-health diagnostics, only if you turn them on (default off). These come from iOS's own MetricKit: crash and hang call stacks with their exception codes, plus aggregate app-health measurements — launch time, hang rate and peak memory. What we do not take is any usage metric MetricKit also offers (launch counts, foreground time, data use): this exists to fix crashes and performance regressions, not to measure how you use the app. Each report carries only your app version, OS version, and device model class (for example "iPhone17,1"), never a device identifier. Critically, these reports are anonymous at ingest: the upload is authenticated so the endpoint is not an open drain, and then your identity is discarded before anything is written. We accept the cost of this (we cannot tell you your crash was fixed) in exchange for holding no crash-linked identity graph.
- Apple's own app analytics, which nothing in our app collects. If you have Apple's analytics sharing switched on (iOS Settings → Privacy & Security → Analytics & Improvements), Apple counts figures such as sessions, crashes and how long people keep the app, and reports them to us in aggregate. We cannot see you individually in it, it is not linked to your account, and — unlike the diagnostics above — it is not ours to switch off. Section 6 describes it.
- Product counters, only if you turn them on (default off): daily counts of a few things members do, such as conversations started, Nows posted and event RSVPs, stored with no account id or other identifier in them, and reported inside the company only when enough people are counted that no one can be picked out. Separately from this switch, we keep daily operational counts, such as sign-ups, active accounts, messages relayed and reports filed, which also carry no identifier.
- IP addresses. Your IP reaches our servers, as it must for any network service. The application does not store raw IP addresses, with one exception: when an upload matches a child sexual abuse hash list, the address it came from is kept with the report we are required to file (Section 3.5). For sign-in security history we store a masked prefix (a /24 for IPv4, a /48 for IPv6) plus a truncated hash of your user agent and either the country our network provider reports for your connection or, where it reports none, that masked prefix, so a returning device is recognizable without retaining the address itself. Rate limiting uses the address transiently and does not persist it. Standard technical logs at our network edge are kept by our hosting and network providers for security and operations (Section 6).
- Oblivious transport. The app contains support for Oblivious HTTP, which hides your IP address from the server that answers only when a relay run by someone else forwards the request. No such relay is in service today, so we make no claim that it hides your IP from us; if an independent operator runs one in the future, we will say so here.
3.8 What we deliberately do not collect
- No analytics or product-analytics SDK. None. The server has zero third-party runtime dependencies. Apple's own aggregate App Store statistics are outside our code; Sections 3.7 and 6 describe them.
- No advertising. There is no ad system, no ad SDK, and no ad-targeting data model in the product, on any tier. Where placement in the app is paid for, it is never ordered by reading your personal data.
- No date of birth. We never ask for one. What we hold about your age is a month and day for the birthday badge, a coarse band for filtering, a numeric age only if you enter one yourself (you choose who can see it), and the estimate the live check makes, which is described in the section "Verification and the live check".
- No contact list, calendar, or photo-library access beyond the specific items you pick.
- No server-side AI on your content. Writing suggestions, reply tone and natural-language search run on your iPhone using Apple's on-device models, and the safety checks in Section 3.5 run on your iPhone too, some of them as simple pattern checks rather than a model. Your content is not sent to us or to any AI provider for these features.
- No face templates, and face checks stay on your device. Face ID and Touch ID stay on your device. The live check compares your face with your own profile photos and estimates your age on your device; we never receive a face template, a face map, a depth scan or a video of you. The one exception is a check held for review: only if you agree at that moment, your device sends one photo from the check and the check's detection results, sealed so that only our review team can open them. The team uses them only for that review, and we delete them no later than 30 days after they arrive. Our biometric data policy is at aroundbackapp.com/biometric-policy.
- No record of which safety or support resources you open. AroundBack does not record which safety or support resources you open. Links to outside services open their websites, and calls or texts you start use your phone's own apps. The list of resources is signed and refreshed once at launch with a request that carries nothing about you.
3.9 Business enquiries (this is not member data)
Our website carries a form for businesses, non-profits and event organizers who want a paid listing in the app. If you use it you are a business contact, not a member, and what we collect is different from everything above. We set it out here because a form that asks a named person for their phone number and a street address deserves a notice at collection whether or not that person is a member.
- What you give us: the name and role of the person enquiring, the organization's legal name, an email address and phone number we can reply on, your city, and a description of what you would like listed. A venue enquiry also carries the venue's street address; an event enquiry, the proposed dates and the venue name; a charity event enquiry, the beneficiary's legal name and EIN, and your confirmation that AroundBack handles no donated funds.
- Please do not put other people's personal details in the description. Tell us about the venue or the night, not about individuals. If you name somebody, we will hold what you typed, and we would rather not.
- What arrives with it: your browser's user agent, and a masked prefix of the source address (a /24 for IPv4, a /48 for IPv6) rather than the address itself. They exist to recognize a flood of automated submissions and nothing else, and they are deleted after 90 days while the enquiry itself stays.
- What we do with it: reply to you, quote and invoice a placement, and keep the commercial record of a listing we published. We do not use these details for marketing lists.
- What we check: for a charity event, a member of our team looks the beneficiary up in public records, such as the IRS Tax Exempt Organization Search, and records what the record showed and who looked. We do not certify anyone's non-profit status.
- Who else sees it: nobody outside the company. Our mail provider is told only that an enquiry arrived, its category and its city, never your details (Section 6). An enquiry is never joined to a member record, never shown to a member, and is not readable by the staff role that operates the venue and event catalog.
- How long we keep it: an enquiry we decline, or that goes no further, is deleted 24 months after we last worked it; an enquiry that became a paid listing is a commercial record and is kept 7 years. Section 8 lists these alongside everything else.
To reach us about an enquiry you sent, write to the address in Section 14. We can tell you what it holds and correct anything that is wrong.
3.10 The launch list (this is not member data)
If you ask us on our website to be told when AroundBack is available, we collect your email address, the day you joined, a two-letter country, and a keyed hash of your IP address, and nothing else about you. You do not need an account, we do not create one, and the list is kept apart from member data and never joined to a member record.
We use it to tell you when the app is available and to send you notices that are important or that the law requires. We never sell it, never share it for anyone else's marketing, and you can remove yourself at any time, from the unsubscribe link in any email we send or by writing to us from that address, which deletes your address.
You confirm your address before we use it. When you enter your address on our website, we send you one message asking you to confirm that it is yours, and we send nothing else until you press the confirm button. If you do not confirm, we delete the address after 24 hours.
Your address is encrypted before it is written down. What sits on our disks is ciphertext, plus a keyed index that lets us recognize a duplicate without comparing addresses in the clear. That protects against a stolen copy of our database or a stolen backup; it does not protect against someone who has taken over our running server, because a running server necessarily holds the key.
Alongside the address we keep the day you joined (the day, not the time) and, where our network provider tells us, the two-letter country the request came from, so that we know where interest is. We also keep a keyed hash of the IP address the request came from, purely so we can recognize a flood of automated submissions; it cannot be read back as an address, and it is deleted after 7 days. Section 8 gives how long we keep the rest.
4. What we cannot read, and what that means
Some limits in this policy are enforced by the architecture rather than by our promise not to look. We list them separately because they are the ones that survive a change of management, an acquisition, or a subpoena:
- We cannot read end-to-end-encrypted message content. We hold no message keys.
- We cannot read chat attachments, private album photos, or private videos.
- We cannot read your Encounter Journal, including the optional backup.
- Nothing we write to disk can locate a member. Every member coordinate that reaches the file is a ~1 km cell center, and the deliberate shares in Section 3.3 are not written to it at all. The only precise coordinates in the file are public venue addresses, and a record that links a member to one of those addresses is kept on a bounded clock, published and enforced, rather than indefinitely; Section 3.3a has the current records and their clocks, generated from our code, along with the limits of the checks behind this. In memory, a coordinate we hold for discovery (your own fix, or a Travel Mode pin) is coarsened after 6 hours; a location you share on purpose stays exact while it is running, because that is what you asked it to do, and is deleted when the share ends.
- We cannot silently swap the encryption key we advertise for a contact: identity keys are published to an append-only transparency log, so a substitution is detectable, and the app lets you compare a safety number with a contact.
- We show no advertising and have no ad system: there is no advertising SDK, ad network or advertising data model, and where placement in the app is paid for (Section 6) it is never ordered by reading your personal data.
We publish a signed statement of what our servers can and cannot see at api.aroundbackapp.com/v1/transparency/capabilities.
Two honest caveats. First, the app checks that the key each chat encrypts to is the key in the transparency log, and refuses to send if they differ. What it cannot do on its own is prove that everyone is shown the same log: phones compare the versions of the log they have seen when they exchange messages, which can reveal a split, but no independent party checks the log today, so a matching safety number is strong evidence but not a proof. Second, an end-to-end-encrypted system still leaks that two accounts communicate. Neither caveat is a reason to distrust the encryption; both are reasons not to overclaim.
5. How we use what we collect
- To run discovery: showing you nearby members as bands, and you to them as a band.
- To deliver messages, calls, and media between devices, without reading them.
- To operate Guardian check-ins and safety escalations.
- To keep the service safe: screening public content, acting on reports, enforcing bans, and resisting one person running many accounts.
- To operate purchases and entitlements.
- To fix crashes, if you opted into diagnostics.
- To meet legal obligations, including mandatory CSAM reporting.
- To respond to your requests and communicate with you about the service.
- To tell people who asked on our website when the app is available, and to send them notices that are important or that the law requires (Section 3.10).
We do not use your data to build advertising profiles, to train AI models, or to sell inferences about you. We do not use sensitive personal information to infer characteristics about you, and we do not process it for any purpose beyond providing the service you asked for, safety and security, and legal compliance.
6. Who else receives data
We share personal data only with the recipients below, only for the purposes described in this policy. Service providers that process personal data on our behalf do so under their terms with us.
Apple. As the platform, Apple necessarily sees:
- Sign in with Apple (Apple knows you signed into our app with your Apple ID; we verify their token against their published keys).
- Push notification delivery (Apple relays our wake hints to your device and, for an incoming call, the caller's display name and account id).
- App Store purchases and subscription notifications.
- A refund you ask Apple for. If you ask Apple to refund an AroundBack purchase, Apple may ask our server a few questions about the purchase. We answer only if you have turned on "Answer Apple's questions about my purchases" in the app. It is off until you turn it on, you do not need it to buy or use anything, and you can turn it off at any time in the same place. When it is on and our server is set up to answer, our reply says whether the purchase was delivered, that a free version of AroundBack was available, that you agreed, and, only where we owe you a refund ourselves, that we would like Apple to grant it. It never asks Apple to refuse a refund, and it contains no name, email address, message, photo or location. We send it only in reply to Apple's request, within the time Apple allows, and only about your own purchase. If the choice is off, or you turn it off before we answer, we send nothing and Apple decides on its own information. An answer already sent cannot be taken back. We keep a copy of what we sent for 30 days. To see or delete the information Apple holds about a refund request, ask Apple directly at privacy.apple.com.
- iCloud, only if you turn on encrypted message history, and then the ciphertext goes to your own iCloud private database. We are not a party to it and cannot read it.
- Apple Maps, when the app shows you a map, looks up a city you type for Travel Mode, or shows a venue's Look Around view: your phone asks Apple for the map, the place or the imagery, as any map app does. If you ask for directions to a venue, the maps app you choose receives the venue's location.
- On-device intelligence, which runs on your phone and sends nothing anywhere.
- App usage, on devices where Apple's analytics sharing is switched on. That is a setting between you and Apple; our app contains no code that produces it.
Apple's handling of that data is described in Apple's own privacy policy.
What Apple gives back to us. Because AroundBack is distributed through the App Store, Apple provides us with aggregate statistics about the app: App Store page views and downloads, active devices, sessions, crashes, deletions, and how long people keep using it. These reports are aggregated by Apple and never identify you to us — we cannot pick an individual member out of them and cannot connect them to an account. Two things we would rather state than let you assume: this data exists because Apple's analytics sharing is switched on, not because our app measures you (there is no analytics SDK in the product); and it is not something we can decline, because it comes with distributing an app through the App Store. Apple's crash reporting reaches us by this same route, separately from the opt-in diagnostics in Section 3.7.
Our own call relay. Voice and video calls are forced through a relay so neither party learns the other's IP address. We run that relay ourselves. It sees both parties' IP addresses while a call is running and relays their encrypted media; it keeps no log of those addresses and cannot decrypt the media.
Our hosting provider and network edge. Our server and its database run on Railway, a hosting provider we rent from, and requests to our API pass through Cloudflare's network on the way to it. Both handle your requests unencrypted at their edge, as any provider that terminates TLS must. They supply the network, the machines and the storage, and we run the software. The database file on Railway's storage is not encrypted by our software: it holds the server-readable data described in Section 3 (account and profile records, coarse location cells, blocks, reports, moderation, consent and purchase records) in a form that a person with administrative access to the volume could read at the file level, alongside the ciphertext we cannot read. Our server also stores public profile photos and voice intros as files, and the ciphertext of chat attachments, on Railway storage. This is a hosting relationship, not a data-sharing one: neither provider is given member data for any purpose of its own, and access to Railway's layer is governed by Railway's terms and security practices. Backup copies that leave the server are encrypted by our software before they leave (AES-256-GCM, under a key that is never stored with the copy or on the volume) and are stored in Cloudflare's R2 storage; Cloudflare does not hold the key. No member data is stored with a third-party analytics, advertising, or data platform.
Our mail provider (Resend). We use a third-party email service. Almost all of it is mail to ourselves: a notification that a business sponsorship enquiry has arrived, a notification that an account access request has arrived, and operational alerts about the health of our own systems. The exceptions are the launch list (people who asked on our website to be told when the app is available; this provider delivers our mail to them), the purchase mail, the fraud-ban safety notice, the re-verification code and the optional email verification code, each described below. Every use is deliberately constrained, and the constraints are enforced in code rather than promised here:
- Company-facing messages carry an internal reference, a category, and a link into our staff console. They do not carry member content, member identifiers, or the contact details of the person who submitted a business enquiry or an account access request.
- The launch-list recipients are people with no account who asked on our website to be told when the app is available. For them the provider receives their email address and nothing else about them; they have no account with us. We use the list to tell them when the app is available and to send notices that are important or that the law requires. Their address is held encrypted. We first send it one message asking the person to confirm it, and we delete it after 24 hours if they have not; once confirmed, it is deleted when they remove themselves or 12 months after they joined, whichever comes first.
- The purchase mail goes to members who buy something, and we send it because the law expects you to receive and keep a record of what you bought. When you buy a subscription or a one-off item, we send a confirmation to the email address on your account (a Sign in with Apple private relay address counts); if your account has no email address, we send nothing. For a yearly plan we also send a notice before it renews and a reminder once a year. For these messages the provider receives your email address, the product, the price and the date from your purchase, and our fixed legal text, and nothing else about you. They carry no advertising, and an email preference does not switch them off, because they are notices the law asks for. They may come from a sender address we use only for purchase mail. If you ask the app to send a copy of your own confirmation again, it goes to the same address.
- The fraud-ban safety notice is one of the four kinds of email we send to members, and the law is why we send it. If we ban a member for fraud or a false identity, New York and Colorado law require us to warn the members who exchanged messages with them (Terms §8). For that single purpose, the provider receives the recipient's email address and the notice text, which names the banned profile's display name, information the recipient already had from messaging them. The notice contains no message content (we cannot read messages; they are end-to-end encrypted), nothing else about the recipient, and nothing about who else received it. No other code path can email a member: a test asserts which paths can address anybody other than us, and the only ones that can address a member are this notice, the purchase mail, the re-verification code and the email verification code.
- The re-verification code for a locked-out member is the second, and we send it only at your own request. If you froze your account, hold no unused recovery code and no Sign in with Apple, and ask us for help (from the signed-out app), a member of our team can start a re-verification. For that single purpose the provider receives the email address you typed into that request, the account ID you typed, a one-time code, and the steps; nothing about your profile, your messages or your face. The code lets you attempt a live check on a new phone; it does not open the account by itself.
- The optional email verification code is the third, also only at your request. If you choose to verify an email address from inside the app, the provider receives the address you typed and a six-digit code, once, to deliver that code to you. We do not store the address: we keep a hash of the code for the five minutes it is valid and, in memory only and for a few minutes, a keyed hash of the address so that one mailbox cannot be sent codes repeatedly; what remains afterwards is a yes-or-no verification mark on your account. The optional phone verification route still has no provider, as described below, and the app is told the code is unavailable rather than that it was sent.
- Accordingly, beyond the purchase mail, the statutory fraud-ban notice, the re-verification code and the email verification code described above, our mail provider receives no member personal data.
If we add any other mail to members or to anyone else, or put personal data in the mail we send ourselves, this section will be updated before that ships.
Legally required recipients. Confirmed child sexual abuse material is reported to NCMEC's CyberTipline as federal law requires, with the associated content and metadata preserved. We may disclose data in response to valid legal process (a subpoena, court order, or search warrant), to protect someone's life or safety in an emergency, or to defend our legal rights, and we will resist overbroad or unlawful demands where we can. We publish a statement of what we are capable of producing, so nobody has to guess.
A copyright claimant, if you answer a copyright notice. If someone sends us a copyright notice about something you posted and you reply with a counter-notice under 17 U.S.C. section 512(g), the law requires us to send a copy of your counter-notice to the person who complained. A counter-notice must include your name, address and telephone number, so they receive those. Beyond what the counter-notice itself contains, we send a claimant nothing else about you, except as valid legal process requires.
Corporate events. If AroundBack is ever acquired, merged, or sells substantially all of its assets, member data would transfer as part of that transaction only under this policy's protections, and we would notify you in the app before any new policy applied to you.
Not enabled today, and named here only so this list stays honest if they are turned on: an identity/age verification vendor, a payment processor for web checkout, the Apple Maps Server API for venue lookups and route estimates, a managed call-media vendor in place of the relay we run ourselves — either a hosted SFU or a hosted TURN provider, since call carriage can move to a third party without the app changing modes — and an SMS provider for the optional phone verification flow (the route exists in the product but has no provider, and says so; if one is enabled, we would receive the number you submit, use it once to send you a code, and not store it; the email verification code goes through the mail provider described above). None of these receives any member data today. If any of them is enabled, this section will be updated before it goes live.
No off-app messaging vendor. We hold no phone number, email address or handle for an off-app Guardian contact. You reach them by sharing a private status link yourself, using whatever app you already use. There is therefore nothing about that person for us to leak, lose, or be compelled to produce.
Venue and event partners. Some venues and events in the app are paid placements, always labeled as sponsored. Partners receive aggregate exposure counts only: how many times their listing was shown in a list and how many times its page was fetched, which are traffic counts rather than counts of people, and how many members are currently marked going. Any count below five is withheld rather than shown. They never receive member identities, a list of who attended, or any per-person data.
Member boosts (Spotlight). A member can buy a Spotlight, which sorts them first inside the distance band they were already in for a fixed window. It is a paid placement, so their card is marked while it runs. The mark carries no timing: other members are shown that a boost is running and never how long is left, because a countdown a stranger could watch drain would reveal when that person paid. Only the member who bought it can see their own remaining time. A boost never changes anyone's distance band, and it never creates a message, a view, or a signal that did not happen.
7. We do not sell your personal data
We do not sell your personal data, and we do not share it for cross-context behavioral advertising or targeted advertising. We have not done so in the preceding 12 months, and we have no mechanism that could. We do not sell or share the personal data of anyone, of any age, and we do not knowingly collect personal data from anyone under 18 (Section 11).
This is not only a policy commitment. There is no advertising system, no advertising SDK, and no advertising data model anywhere in the product; there is no analytics vendor in the app or on our servers (the aggregate App Store statistics Apple gives us, described in Section 6, come from Apple and not from anything we run); and the server carries no third-party runtime dependencies through which such a transfer could occur.
Because we do not sell personal data or share it for targeted advertising, there is no "sale" or "sharing" for you to opt out of, and no processing that a universal opt-out signal (such as Global Privacy Control) would change. If our practices ever change, we will update this policy first, provide the required opt-out rights, and honor recognized opt-out preference signals from that point on.
A separate, explicit no-sale statement covering consumer health data appears in the Consumer Health Data Privacy Policy.
8. How long we keep things
We keep personal data only as long as it is needed for the purposes in Section 5, and the product is built so that most of it expires on a short, enforced clock. Where a fixed period exists in the software, it is stated below. Where no fixed period is possible, we state the criteria that decide the retention instead; in every such case the data is deleted when you delete your account, except for the records identified below as outliving it, which Section 9.4 lists together.
| Data | Retention |
|---|---|
| Precise coordinate held for discovery, in memory | Coarsened to a ~1 km cell center 6 hours after the fix. The same rule and the same clock apply to a Travel Mode pin; the pin itself is dropped entirely after 30 days if you never change it |
| Precise coordinate that could locate a member, on disk | Never written: a discovery fix and a Travel Mode pin are coarsened to a cell center before the store is serialized; a Guardian live fix and an On My Way fix are omitted from the file entirely. (Precise venue addresses in the public directory are stored; they are not member locations. What links you to one of those addresses, and for how long, is Section 3.3a) |
| A record that links you to a venue in our directory | Kept on a published clock and then deleted; see the table in Section 3.3a, which is generated from the same code our deletion sweeps run on. Deleting your account removes every record in that table at once |
| Message ciphertext | 30 days by default; less if you set a disappearing timer; deleted 5 seconds after the recipient acknowledges delivery |
| Chat attachment ciphertext | 30 days, or on burn/erasure |
| Conversation record and "who has messaged whom" edge | Forgotten after 90 days of silence with no live messages (or 3× the message retention, whichever is longer) |
| Curtain (discreet Now) threads | Destroyed with the drop; hard cap 12 hours |
| Now drops | Your chosen duration, capped at 12 hours |
| Guardian encounters | 24 hours after the encounter ends. A check-in that escalated and was never resolved has no end time, so it is deleted 72 hours after the alert. A check-in ended with your duress code is deleted on the ordinary 24-hour clock, measured from the end time your own screen displays, so that it cannot be told apart from a genuine safe return by how long it lingers |
| Guardian live coordinate | Deleted the moment you mark safe, cancel, or switch sharing off; the record outlives it, the position does not. Under duress it deliberately keeps flowing |
| Guardian contact alert and Lock Screen tokens | Deleted when the check-in ends or its link is reset, when the contact turns alerts off, and when the Lock Screen display ends; a token resting only on the App Clip's temporary permission 8 hours after it was registered; a Lock Screen display token at the latest 8 hours after the display began. A member's own display token also goes if either person blocks the other, the member deletes their account, or signs out of that phone |
| On My Way trips | Capped at 2 hours; auto-ends once you have arrived; visible for about an hour after it ends, then dropped. Your live position is deleted when the trip ends, and in any case within about 5 minutes of the last fix |
| Ghost mode | Your chosen window (timed windows capped at 30 days), or until you turn it off |
| Plan change notices (that one of your event plans changed; never which event) | Deleted when you dismiss it, and at the latest 14 days after the most recent change it tells you about, and never later than the last of the RSVPs or saves it replaced would itself have been deleted. A marker that we sent you a plan change push, kept so that we send at most one a day, is deleted 24 hours after that push |
| Sessions | 30 days |
| Sign-in security history | Most recent 50 events per account (masked IP prefix + context hash only) |
| Sign in with Apple refresh token | Until you delete your account. If Apple cannot be reached at that moment, the encrypted token, holding no account identifier, is retried for up to 30 days and then discarded |
| Suspensions | Purged when the term lapses. Bans persist until lifted, and we retain the minimum hashed identifiers needed to prevent a banned person from returning; a granted appeal releases them |
| CSAM preservation records | One year after NCMEC acknowledges the report, isolated, as required by law, and longer only while legal process or a legal hold applies; then destroyed, leaving only the file's digital fingerprint, the report number and the dates. Not removed by an erasure request |
| Sealed record of legal holds on preserved child safety material (which preserved file was put on hold or released, by internal reference number; which of our staff did it, when, and the short reason they typed; we instruct our staff never to put pictures, member names or contact details in it) | Kept with no end date, including after a hold is released and after the preserved file itself has been destroyed at the end of its year. A copy is stored away from our servers, encrypted so that only we can open it. An erasure request does not remove it, because the law requires us to keep what it protects. |
| Moderation cases, reports, appeals, audit log | Retained while your account is active and for as long as needed after a case closes to enforce our policies, defend legal claims, meet legal obligations, and maintain repeat-offender history. On account deletion these records are severed from your identity (tombstoned) rather than deleted |
| Public profile photos and voice intros | Until you remove them, a moderator removes them, or your account is deleted, whichever comes first |
| Anti-abuse risk signals and device↔account links | Retained while your account is active, as our defense against one person running many accounts; deleted with your account, except the hashed ban-evasion anchors described above for banned accounts |
| Visitors (who viewed your profile) | Retained while your account is active; cleared on account deletion |
| Purchase and subscription records | Retained while your account is active; deleted with your account, except a one-way hash of each App Store transaction, kept so the same purchase cannot be claimed by a second account. Apple keeps its own record of your purchases |
| Your choice about answering Apple's refund questions | Kept, with when you made it and which wording you saw, while your account is active; deleted with your account |
| Crash/hang telemetry (anonymous) | 30 days, capped at the most recent 200 reports plus up to 500 aggregate signature counters; never linked to you |
| Key Transparency log entries | Append-only and not removable by design: each entry holds an identity public key and a handle derived from an account id rather than the account id itself, and removability would defeat the security property it exists to provide. When you delete your account we stop serving your key to anyone and add an entry marking it retired. It contains no profile content, no location, and no messages |
| Encrypted journal backup and saved-phrases backup | Until you delete them or delete your account |
| Fraud-ban notice records (who we notified after a fraud ban, when, and the outcome) | 36 months, proof of compliance with dating-service fraud-ban notice laws; the banned profile's display-name snapshot ages out with the record |
| Consent records | Your current answer for each purpose, until you delete your account; an answer you later changed, 24 months after you changed it. Deleting your account deletes them, leaving only a count that names nobody |
| The Hype | A confirmed meeting: 8 days, then a count that names nobody. An endorsement: until it is withdrawn (then 30 days) or either account is deleted |
| Takedown request details (who you are and how to reach you) | 90 days after we have dealt with your request, or sooner if you delete your account |
| Takedown compliance record (names nobody) | Kept, with no end date |
| Copyright notices, counter-notices, strikes and our records of what we did about them, including the names and contact details of the person who sent a notice and of the member it concerns | 3 years after the matter closes, the time within which a copyright claim can be brought (17 U.S.C. 507(b)), and then deleted. Deleting your account does not delete them sooner (Section 9.4) |
| A photo or voice intro we hid after a copyright notice | Kept hidden and sealed, not destroyed, until the time the law allows the person who complained to start a court action has ended: 10 to 14 business days after we receive your counter-notice, so at most about 3 weeks. Then we put it back, unless that person has told us they started a court action, in which case it stays down. If no counter-notice arrives, we keep the file hidden and sealed for 90 days after the notice and then delete it. |
| Everything else tied to your account | Until you delete your account |
| Business sponsorship enquiry that goes no further | 24 months after we last worked it. This is a business contact record, not member data |
| Business sponsorship enquiry that became a paid listing | 7 years, as the commercial record of a transaction |
| Masked source prefix and user agent recorded with an enquiry | 90 days, after which they are deleted and the enquiry itself stays |
| Launch-list address you have not confirmed | Deleted after 24 hours |
| Launch-list address you have confirmed | Kept until you remove yourself, and in any case no longer than 12 months from the day you joined |
| Launch-list abuse hash | The keyed hash of the source IP is deleted after 7 days |
| Account access request (the account ID and email you typed, and your words, when you ask for help getting back into a frozen account) | 24 months after we last worked it, whether it led to a re-verification or not |
| Masked source prefix and user agent recorded with an account access request | 90 days, after which they are deleted and the request itself stays |
| A re-verification of a frozen account (its state, timestamps and outcome; never the code, which is stored only as a hash) | 14 days after it ends; the audit entry that a member of our team started it outlives it in the verification audit log |
| Live check result (the signed result of each check, with the age estimate and the consent version) | Every passed check and the most recent 20 failed checks, while your account is active. After you delete your account, 90 days with your account ID removed, then deleted; longer only if the law requires us to preserve the account's records |
| Live check session (the record that a check began) | Deleted 24 hours after it expires |
| A held check's photo and results (only if you chose to share them) | Deleted when your review is decided, at once if you delete your account, and no later than 30 days after they arrive. Never in our backups |
| Age decision records (an appeal of an age block, and the log of age decisions) | Deleted one year after the case closes. Kept for as long as an age block on the account stands, and longer only if the law requires us to preserve the account's records |
| Operational backups | Local copies roll over after about 14 days, and a deleted account ages out of them as they rotate. Encrypted copies kept off our server, for recovery if its disk is lost, are stored with Cloudflare (Section 6) and have no fixed expiry, so an account deleted after a copy was made remains in that copy, sealed under a key Cloudflare does not hold. Backups are never used to resurrect deleted data |
| Edge logs | Kept by our hosting and network providers on their own schedules (Section 6) |
9. Your rights and your controls
9.1 In the app, free, on every tier, without contacting anyone
- Export everything. One request returns your data from every subsystem that holds any: profile, your own last submitted coordinate, signals sent and received, Now sessions, saved items, safety settings, purchases, encrypted backups, the reports and appeals you filed, and the rest. It does not include reports other members filed about you or the moderation case they sit on: those are our records of how we handled a report, and an export that carried them would tell you who reported you. You can ask us whether such a report exists. A few records come as a summary rather than a copy: your anti-abuse risk level rather than the signals behind it, how many people viewed your profile rather than who, how many Guardian check-ins you started or guarded rather than their details, and references to your public photos and voice intro rather than the files themselves. Because a full export is a high-value target, it requires a recent re-authentication: a stolen session token alone cannot pull it.
- Delete your account. Also gated on recent re-authentication and an explicit confirmation, because it is irreversible. Deletion fans out across every module that holds your data and revokes all your sessions. It is not paywalled and never will be; nor are blocking, reporting, hiding, pausing, or panicking. Deletion works even if your account is suspended or banned.
- Correct your information. Everything we hold that you entered (your profile, your settings) is editable in the app directly.
- Withdraw consent. The consents you gave at onboarding each have a switch in You, App & data, Privacy choices, and withdrawing is as easy as giving. For your sensitive profile fields and your precise location, withdrawing removes the data that consent covered — we delete what we hold, not merely stop using it. Product counters hold nothing of yours to remove: they are counts with no identifier in them, so withdrawing simply stops your actions being added to them. The operational counts in Section 3.7 are not affected by this switch. Crash diagnostics are the one exception, and the reason is itself a privacy choice. Crash reports are stored with nothing recording whose they are, so there is no way to find yours and delete them without first creating exactly the thing we chose not to keep: a record linking you to your crashes. Withdrawing stops the next upload; the reports already held carry no link to you and age out on their own retention clock.
- Control your visibility: Anonymous mode, hide distance, pause, incognito browsing, Ghost mode, per-person Now excludes, per-field profile visibility, and emergency hide.
- Control your messages: disappearing timers globally or per conversation, unsend, typing indicators (off by default), and read receipts.
- Control your journal: it is on-device by default; the encrypted backup is opt-in; and one tap destroys both the encrypted file and its key.
9.2 Your rights under US state privacy laws
Depending on where you live, state law gives you some or all of the following rights over your personal data. We honor these rights for every member, in every state, rather than making you work out which statute covers you:
- Right to know / access: confirm whether we process your personal data and receive a copy of it.
- Right to correct inaccurate personal data.
- Right to delete your personal data.
- Right to portability: receive your data in a portable, machine-readable format (the in-app export).
- Right to opt out of the sale of personal data, sharing/processing for targeted advertising, and profiling in furtherance of decisions that produce legal or similarly significant effects. We do not do any of these things, so there is nothing to opt out of; if that ever changes, we will provide the opt-out first (see Section 7).
- Right to limit use of sensitive personal information (California). We use sensitive personal information only to provide the service you asked for and for the safety, security, and legal-compliance purposes California's regulations permit; we do not use it to infer characteristics. Because our use stays within those permitted purposes, California law does not require a "Limit the Use of My Sensitive Personal Information" link, and there is no additional limiting for the right to effect. If our use ever expands, we will post the link first.
- Right to non-discrimination / non-retaliation: we will not deny you the service, charge you a different price, or degrade your experience because you exercised a privacy right. Safety features, data rights, and cancellation are free on every tier by design.
- Right to appeal (most states): if we refuse to act on your request, you may appeal, and if the appeal fails you may contact your state Attorney General.
How to exercise these rights. Use the in-app tools (correct your profile: You → Edit profile; withdraw a consent: You → App & data → Privacy choices; export your data or delete your account: You → Account & security → Passkeys, sessions, recovery & data → Your data; confirm a request you emailed us: You → Account & security → Privacy requests), or email privacy@aroundbackapp.com from any address; you do not need an account to write to us. We will verify your identity before acting on access, correction, or deletion requests. Because the in-app tools are authenticated, they are the fastest path; for email requests about an account, we verify by asking you to confirm control of the account through a signed in-app prompt, because we deliberately hold no shared secret or knowledge factor we could quiz you on to verify you (no password, no date of birth, possibly a relay email), so knowledge-based verification is not a safe option. If your request is about a business enquiry rather than an account, the in-app prompt is not available to you and is not required: write to us from the email address on the enquiry, and replying from that address is how we verify you. The same is true of the launch list: you have no account to prompt, so the unsubscribe link in any email we send takes you off it immediately without involving us at all, and writing to us from the address itself works too. If you no longer control the address, tell us and we will find another way to be satisfied that the request is yours. If we cannot verify a request, we will tell you why. Authorized agents (California and other states) may submit requests on your behalf with proof of your written permission; we will still confirm the request with you directly through the app.
Timing. We respond within 45 days, extendable once by another 45 days where reasonably necessary (we will tell you if so). Appeals: reply to our decision or write to privacy@aroundbackapp.com with "Appeal" in the subject; a different reviewer decides within 45 days and we explain the outcome in writing, including how to reach your state Attorney General if you remain unsatisfied.
9.3 California-specific disclosures
The California Consumer Privacy Act asks us to map this policy to its categories. In the preceding 12 months we have collected the following categories of personal information: identifiers (account id, Apple subject identifier, email address, display name, and the email address of someone on the launch list); customer records (subscription entitlement state); protected classifications insofar as optional profile fields you choose to complete reveal them (for example sexual orientation, gender identity or ethnicity); commercial information (purchase and entitlement records); audio and visual information (profile photos, voice intros, and end-to-end-encrypted media we cannot read); geolocation data, including precise geolocation (as described in Section 3.3); internet or electronic network activity limited to service interactions (presence tier, profile visits, Signals sent and received, sign-in security history); professional or employment-related information (the name, role and business contact details of someone who sends a business enquiry, Section 3.9); sensitive personal information (precise geolocation; profile fields revealing sexual orientation, gender identity or racial or ethnic origin; message content we relay but cannot read). Sources: you, your device, Apple, other members who report you (Section 3), and the people who fill in our business enquiry, launch-list and takedown forms. Purposes: Section 5. Categories of recipients: Section 6. Retention: Section 8.
We have not sold or shared personal information (including that of minors, whom we do not knowingly serve at all) in the preceding 12 months, we use and disclose sensitive personal information only for purposes permitted by section 7027(m) of the CCPA regulations, and we offer no financial incentives in exchange for personal information.
Shine the Light (Civil Code section 1798.83): we do not disclose personal information to third parties for their own direct marketing, so there is nothing to request under that statute.
Do Not Track and opt-out preference signals. Our app and website do not track you across other companies' sites or apps, and no third party collects personal information across sites or over time through our service. Because there is no tracking, cross-context sharing, or sale, browser Do Not Track and Global Privacy Control signals have nothing to change; we treat every visitor as if they had sent one. If our practices ever change, we will honor recognized opt-out preference signals as the law requires and say so here.
9.4 What deletion cannot reach
Stated plainly rather than buried: records we are legally required to preserve (CSAM preservation under federal law); the append-only key transparency log (Section 8); hashed ban-evasion anchors for banned accounts; moderation cases, reports and appeals, which stay with your identity removed; copyright notices, counter-notices, strikes and our records of what we did about them, which keep the names and contact details of the person who sent a notice and of the member it concerns for 3 years after the matter closes (Section 8); if you were banned for fraud, the record of the warnings we sent, which keeps your profile's display name for up to 36 months; a one-way hash of each App Store purchase, kept so the same purchase cannot be claimed twice; aggregate counts that contain nothing about you individually; operational backups: the local copies until they rotate out, and the encrypted off-site copies, which have no fixed expiry (Section 8); an encrypted message history you turned on in your own iCloud, which stays there unless you switch it off in You, Messages, iCloud backup before you delete your account or remove it in iCloud; and any data already delivered to another member's device: a message you sent that someone else received is on their phone, and we cannot reach into it.
10. Security
Your account uses Sign in with Apple and/or passkeys. We do not offer passwords, so there is no password of yours to leak.
On your device: the app can be locked with Face ID or Touch ID, or an unlock PIN you set for the app; the Encounter Journal is encrypted at rest under a key that exists only on that device and is unavailable while the device is locked; message history and the search index are encrypted; and a duress mode presents a decoy account that reads and writes nothing real and shows nothing of the real account, whose data stays on the phone.
In transit and at rest: TLS everywhere, and end-to-end encryption for messages, calls and private media.
We also detect and warn about a compromised device (jailbreak, debugger, code injection). We warn once and never block; locking someone out of a safety app because their phone looks unusual is the wrong trade.
We maintain a written security program with administrative, technical, and physical safeguards appropriate to our size and the sensitivity of the data we hold. No system is perfectly secure, and we do not claim otherwise. If a breach affects you, we will notify you and regulators as the law requires, without unreasonable delay. Security reports: privacy@aroundbackapp.com.
11. Children
AroundBack is for adults only. You must be 18 or older to create an account, and 19 or older if you live in British Columbia, New Brunswick, Newfoundland and Labrador, Nova Scotia, the Northwest Territories, Nunavut or Yukon. At sign-up the app asks Apple's Declared Age Range for your age range, and where your device cannot give one, you confirm that you are 18 or over yourself; we do not claim to have independently verified every member's age. We do not knowingly collect personal data from anyone under 18. If we learn an account belongs to someone under 18 (through a report, a device signal, or any other source) we block it, and the holder can ask a person to review the block. To report an underage account: use in-app reporting or email support@aroundbackapp.com.
12. International transfers
The app is offered in the United States and Canada. We do not serve residents of the EU, UK or other jurisdictions as members.
The business inquiry form is different. It is a web page, it is reachable from anywhere, and a venue or an event organizer outside the United States is welcome to use it. If you do, the business contact information you submit is transferred to and processed in the United States. Section 12.1 is the notice that goes with that.
12.1 If you submit a business inquiry from the EU or the UK
This section applies only to the sponsorship inquiry form. It does not describe membership, because the app is not offered where the UK GDPR or the EU GDPR apply.
- Who is responsible. Maison Mason LLC d/b/a AroundBack, at the address in Section 14, is the controller of the information you submit. We have not appointed an Article 27 representative, because our only processing of EU or UK personal data is this form; if that changes we will appoint one and say so here.
- Why we hold it and on what basis. To read your inquiry, reply to it, and if we go ahead, arrange and run the placement. Our lawful basis is our legitimate interest in responding to a business approach that you made to us. We weighed it: you contacted us in a professional capacity, you chose what to send, the information is about a business rather than your private life, and you can ask us to stop at any time. Where we go on to agree a placement, we also rely on performance of a contract with your organization.
- What we do with it. Nothing but the above. We do not use it for marketing, we do not build a profile from it, and there is no automated decision-making.
- Who receives it. Our own staff, with administrative access; our hosting provider and network edge, described in Section 6; our mail provider, which is told only that an inquiry arrived, its category and its city (Section 3.9); anyone we are legally required to disclose to; and, on a corporate event, an acquirer under this policy's protections, as Section 6 describes.
- How long. As set out in Section 8: 24 months for an inquiry that goes no further, seven years for one that became a live listing, and 90 days for the anti-spam fields.
- Where it goes. To the United States. We rely on the UK International Data Transfer Agreement and the EU Standard Contractual Clauses for onward transfers to processors, together with the technical and organizational measures in Section 10.
- Your rights. Access, rectification, erasure, restriction, portability, and objection, including objection to processing based on legitimate interests. Use the route in Section 9.2; you do not need an account.
- Complaints. You may complain to your supervisory authority: in the UK the Information Commissioner's Office, and in the EU the authority for the country you live in.
- Is it required? No. Nothing obliges you to send us an inquiry, and there is no consequence to not sending one beyond our not knowing you would like a listing.
A note on other people's details. This notice covers the information you type. We ask you on the form not to include other people's personal details, and if you do include them anyway, write to us and we will remove them.
13. Changes to this policy
We will post changes here with a new "last updated" date. For material changes, we will notify you in the app before they take effect and, where the change concerns a consent you gave, ask again rather than assume. The effective date at the top of this policy tells you which version you are reading.
14. Contact
Maison Mason LLC d/b/a AroundBack
Legal Affairs
2248 Broadway #1231
New York, NY 10024
Privacy requests and questions: privacy@aroundbackapp.com
Support: support@aroundbackapp.com