Legal
Privacy Policy
Last updated 16 November 2026. This policy explains what GrailFox collects, why, and who else touches it.
1. Who we are
GrailFox is made and run by Maksim Larkin, a sole trader in Berlin, Germany, who is the controller for the personal data described here. The full postal address is on the Impressum. For anything in this policy, write to support@grailfox.net.
2. What we collect
GrailFox has no sign-up, so we hold no name, email address or password unless you write to us. In normal use the app processes four kinds of data: the photograph you take of a card, the records behind your account — your collection, your scan history and your remaining scan allowance — the purchase and subscription state of that account, and product analytics describing how the app is used.
The analytics are a fixed list of events about what happens in the app — it opened, a scan started, succeeded or failed, the paywall was shown, a purchase completed. Each carries the same random account identifier, the app and build version, which build it is, and your device model, manufacturer and operating system version. Our analytics provider also derives an approximate location — country, region and city — from the network address the request arrives on, although the app never asks for your location and cannot read it. No event carries a photograph, a card, a name or an email address — except what you type yourself into the one line of text described in the next paragraph — and none uses your device's advertising identifier. From app version 1.0.8, if the app counts you as being in the European Economic Area, it sends them only if you switch them on; §5 explains how it decides, and what it cannot tell.
From app version 1.0.9, when a paywall is shown the app also records which plans it showed, whether their prices loaded from the store, which plan's free trial it offered, if any, and which plan was selected before you tapped anything; and every event says whether your account began on version 1.0.9 or later. The first time you leave a paywall without buying, the app may ask you, once, what stopped you. Answering is optional: you can pick one of five fixed answers or skip the question, and after “Something else” you can add a line of text of up to 120 characters. Either way the result — your answer, or that you skipped — is sent as one more analytics event, with the same random account identifier, and we use it as we use the rest of the analytics: to understand where the app fails the people who try it. The line of text is sent as you typed it, so leave out your name, your email address and anything else personal. Where the app asks before sending analytics, as §5 describes, it asks this question only once you have switched them on.
If the app hits an error it did not expect, it also sends a diagnostic report: what went wrong and where in our code it happened. Before it leaves your phone we strip the file paths, shorten the message, and remove the contents of any variable involved, so what we receive tells us which line failed and nothing about you or your cards.
Your account is tied to a random identifier created on first launch, and to nothing else about you. That identifier is kept on your device and is the only way back into the account, so a phone that is lost or reset loses access to the collection.
Your scan history is one of those account records, and it is worth spelling out: we keep a record of every scan attempt — what the app answered, whether it charged an allowance, whether you corrected the answer, how long it took, and the technical measurements of the photograph that §3 lists — including attempts that failed. A failed scan is recorded but never charged; the record of a failure is what tells us recognition is going wrong before the reviews do. The record never contains the photograph or your name, because we hold neither.
Like any server, ours also sees the technical connection data every request carries — your IP address, the time, the request line, the page you came from and your browser's user-agent string — which lands in the web server's log and nowhere else; we use that log to keep the site running and to count the visits each page and campaign brings. The link grailfox.net/get reads the last two, and the operating-system name Chrome and other Chromium browsers send with each request, to send your phone to its own app store and to tell that store which page or campaign sent you; the tag it adds names the source, never you.
3. Card photographs
We never keep a copy of your photographs. Each photograph reaches our server and is passed on, without your identifier, to OpenAI, our recognition provider, which reads the card's name, number and printing clues off it and returns the identification; we put it in no database and write it to no log, and nothing of the image is left once the request has finished. What we keep is what was read off the card, as part of the scan record §2 describes, and a few technical measurements of the photograph, kept with that record: its dimensions, file size and file format; how sharp and bright it is, its contrast, and how much of it is very bright or very dark; how long measuring it took; and which recognition model we asked to read it and how much work that took. Handling an upload of that size means the web server and the framework spool it to a temporary file for the seconds the request lasts — that is how any upload of a few megabytes is received — and the system removes that file when the request ends. When you add a card to your collection, the app keeps that photograph on your device so your collection shows the card you actually own; the app deletes it when you remove the card, deleting GrailFox takes the copies on your phone with it, and it is included in your device backup if you use one — a backup you have already made keeps whatever it captured. While you scan a series, the app also keeps the photos you take on your device, to send them and to show them in the series, and deletes them when the series ends: at once when you discard it or when an add leaves nothing in it, and, if it expires after seven days without use, the next time the app opens and finds it gone. An add that leaves anything in the series keeps all its photos, those of the cards just added included, until the series ends. While they are kept, they are included in your device backup in the same way.
The photograph cannot be rebuilt from those measurements. They are part of your scan history, so like the rest of it they are tied to your account's random identifier, and they are kept for as long as that history is (§6); we keep them to find out why scans fail. They are recorded by our server, with the scan's record, not sent by the app, so they are not the product analytics described in §2 and §5, and the “Share usage statistics” switch does not turn them off.
Under our current settings OpenAI may also use photographs to develop and improve its services, including training its models. We have no special deletion arrangement with OpenAI, so how long it keeps a photograph is governed by OpenAI's own policies. Keep the frame on the card: whatever is in the photograph makes that journey with it.
4. Third parties
Seven companies help deliver the service. Each is listed with what it receives; none receives more.
OpenAI (United States) identifies the card from the photograph, as §3 describes. OpenAI is certified under the EU-US Data Privacy Framework, which the European Commission recognises as adequate protection for transfers to certified US companies.
JustTCG supplies the card catalog and the price estimates. Our server asks it about the card, never about you — no personal data is sent to it.
Apple handles payment, subscriptions and app distribution on iPhone. Your purchase there is a purchase from Apple, made under Apple's own terms and privacy policy; we never see your payment details.
Google handles payment, subscriptions and app distribution on Android, through Google Play. Purchases there are processed by Google under Google Play's own terms and privacy policy; we never see your payment details.
RevenueCat (United States) manages purchase and subscription state for us. It receives the random account identifier, purchase history and device identifiers, and processes them on our behalf; transfers rest on the EU Standard Contractual Clauses incorporated in its data-processing terms.
PostHog receives the product analytics described in §2, on its European service with the data held in Frankfurt. It receives the random account identifier, the app version and device details, the approximate location it derives from the network address, the error diagnostics described in §2, and, from app version 1.0.9, your answer to the question §2 describes, or that you skipped it — never a photograph or a card, and never a name unless you type one into that answer.
Because both are keyed to the same random identifier, RevenueCat also sends our purchase events directly to PostHog — the product bought, the amount, the currency and the store — without them passing through the app.
Hetzner hosts our server, in Germany.
This website loads no third-party scripts, fonts, images or analytics, and sets no cookies.
5. Legal bases
Where the GDPR applies: we process the photograph, your collection and your purchase state to provide the service you asked for (Art. 6(1)(b)). Six things rest on our legitimate interests (Art. 6(1)(f)): the server's connection logs, in keeping the service running and secure and in counting the visits each page and campaign brings; the store link's reading of your browser's user-agent, its operating-system hint and the page you came from, in sending you to the right app store and telling it where you came from; the scan records described in §2, with the photograph measurements described in §3, in knowing whether the identifications we give are right and why scans fail — the rate at which users correct us is the only honest measure of the first; the product analytics outside the European Economic Area, in understanding which parts of the app people reach and where it fails them; the error diagnostics described in §2, outside the European Economic Area, in finding and fixing what breaks; and the purchase events RevenueCat sends to PostHog (§4), in knowing what is bought — these reach PostHog wherever you are and whatever you choose below. You can object to any processing that rests on a legitimate interest (Art. 21) by writing to us. Purchase records are kept because tax and accounting law requires it (Art. 6(1)(c)). GrailFox makes no automated decisions about you with legal or similarly significant effect within the meaning of Art. 22 — identifying a card is not a decision about you.
In the European Economic Area — the EU, Iceland, Liechtenstein and Norway — the product analytics and error diagnostics rest on your consent instead (Art. 6(1)(a) GDPR, and Art. 5(3) of the ePrivacy Directive — in Germany, §25 TDDDG — for what the app stores on your phone), on app version 1.0.8 or later and where the app counts you as being there. It counts you as there when your phone's region (on Android, the country of the phone's first language) is one of those countries, or, from the moment it first reads prices from your App Store or Google Play account, when that account is in one. From then on it collects no analytics or error diagnostics until you switch on “Share usage statistics”: the switch is offered on the screen that asks, before you scan, whether your photos may be sent to OpenAI, and in the app's Settings under Privacy. It is off until you turn it on, and leaving it off changes nothing else — scanning works the same. The switch covers only what the app sends; the scan records of §2, with the photograph measurements of §3, are kept on our server whichever way it is set, on the legitimate interest above. When the app starts while the switch is off, it deletes any analytics data left on your phone. Once it is on, the app sends what §2 lists. Those events are pseudonymous, not anonymous: each one is tied to the random identifier the app created for your account, never to your name or email address. You can switch it off again at any time in Settings, and collection stops from that moment: the app sends what it had already collected and one last event recording the change, and nothing after it. Switching off does not undo what was sent before; §6 says how long that is kept. Under Art. 8 GDPR you can give this consent yourself from the age of 16, or from a lower age where your country sets one — never lower than 13. If you are younger, leave the switch off.
Some phones in the European Economic Area send analytics and error diagnostics without being asked. Earlier versions of the app, on iPhone and on Android alike, do not ask, and keep sending both until they are updated; Google Play has offered the Android app there since October 2026. A phone whose region setting and app-store account are both outside the area looks to the app like a phone used elsewhere, so it is not asked either. And on a phone whose region is outside the area but whose App Store or Google Play account is in it, what the app collects before it first reads prices from that store can still be sent. If any of this is you, write to support@grailfox.net to object, or to have those events deleted; §7 explains how we try to find them, and why we may not be able to.
6. Retention
We keep no photographs at all — the copies described in §3 are on your device, not ours; the app deletes one when you remove its card, deletes a series' photos when the series ends, as §3 describes, and deleting GrailFox takes the ones on your phone with it — a backup you have already made keeps whatever it captured, as §3 says. What OpenAI keeps is governed by its own policies, as §3 says. Your collection and scan history, the photograph measurements of §3 included, are kept for as long as your account exists — we cannot tell an inactive account from a patient one, so that means until you ask us to delete it. The web server's access logs are deleted after 14 days. Analytics events are kept for as long as they are useful for the purpose in §5 and are deleted when they are not; we will state a fixed period here once we have one rather than name a number we do not yet keep to. Purchase records are kept for as long as tax and accounting law requires.
7. Your rights
You may ask for access to, correction of, or deletion of personal data we hold, ask for a copy in a portable form, and object to or ask us to restrict processing (Art. 15–21 GDPR). One honest caveat: your account carries no name or email, so we cannot find it from your message alone (Art. 11 GDPR). Write to support@grailfox.net and we will work out with you whether your account can be identified; removing individual cards is something you can do directly in the app. You can also complain to a data protection supervisory authority — ours is the Berliner Beauftragte für Datenschutz und Informationsfreiheit in Berlin.
8. Children
GrailFox is not directed at children under 13, and we do not knowingly collect their personal data.
9. Changes and contact
If this policy changes materially — a new processor, a new purpose — we will update the date at the top and note the change in the app's release notes before the change takes effect. Questions go to support@grailfox.net, or to the postal address on the Impressum.