Privacy Policy: Culinoria
Last updated: 2026-11-01
1. Data Controller
Christopher Hader Am Aignerteich 11 2353 Guntramsdorf Austria Email: support@culinoria.com Phone: +43 681 20705682
2. Overview
Culinoria is an iOS recipe management app. Recipe data is stored locally and synced via the user's private iCloud account; the provider has no access to it. No user accounts, no tracking, no advertising/analytics SDKs.
When an AI import feature is actively used, data is transmitted to the services listed in section 4. Independently of that, the server also stores a cross-device identifier per household plus usage/cost counters for quota and subscription checks (section 5) - this is a pseudonymous, persistent identifier, not an anonymous one.
3. Local data processing and iCloud sync
Recipes, ingredients, preparation steps, photos, ratings, cooking history, meal plan, and shopping list are stored locally using Core Data and synced via the private CloudKit database of the Apple account (operator: Apple Inc.). The provider has no access; sync can be disabled in iOS Settings.
Deleting the app removes local data; iCloud copies remain until removed there separately (see section 11 on iOS backups). The server-side data described in section 5 is NOT covered by this.
Share feature (Share Extension)
Handing off links and photos from other apps (shared text is only reduced to the link it contains) runs via a local App Group container between Culinoria and its share extension; no data is sent to third parties. The source stays on the device until explicit completion or cancellation, including across restarts. Transmission per section 4 only occurs after explicit confirmation of the import in the app.
Passing recipes on as a file
On explicit request, "Share Recipe as File" creates a file (.culinoria) containing a single recipe: title, ingredients, steps, tags, notes, source reference, and photos. Ratings, cooking history, favorite status, and household, shopping list, and meal plan data are not included. The file is unencrypted and is handed over via the iOS share sheet; where it ends up (e.g. AirDrop, Messages, email, Files) is decided solely by the user. This is an action outside the app and outside iCloud: anyone who receives the file can read its contents. The provider does not receive the file. When such a file is opened, the app reads it locally, shows a preview, and saves the recipe as a new recipe only after confirmation; nothing is sent to servers and no AI call is made.
4. AI recipe imports
These features are optional; the app is fully usable without them. Before first using one of these features, the app shows a notice naming the services involved (see "Legal basis" below).
- Photo import (camera, photo library): downsized photos as image files; when scanning with the camera document scanner or choosing from the photo library, up to 6 photos or pages of one recipe in a single request
- File import (PDF, image file): if the PDF contains enough text, the text read from it (at most 20,000 characters); otherwise up to the first 6 pages as downsized images. Image files are sent like photos
- Web link import: entered URL; server fetches the page, extracts text and, if available, a cover photo (schema.org markup)
- YouTube import: entered URL; title, description, thumbnail and, where available, a comment by the channel owner via the YouTube Data API (fetched by our server)
- TikTok/Instagram/Pinterest import: entered URL. The app loads the public page of the post (for TikTok, as a fallback, the public oEmbed endpoint) directly from the platform on your device and reads the description (caption/title) and the preview image address from it; the platform sees your IP address and the link. Only the text that was read (with the name of the platform) is sent to our server and Anthropic, not the URL
- Description pasted by you (fallback for video links): the text you paste yourself; like a text that was read, it goes to our server and from there to Anthropic
The app reads photos from the photo library and files only after you have chosen them in the system picker; access to your whole photo library is not requested. The selection is not stored, and the file stays where it was. Several images or pages of one recipe are sent to the server in a single request and count as one import.
All import paths also transmit the selected target language, the unit system and the category names of the shopping list (default or self-defined categories; used during import to assign ingredients to categories), which are passed on to Anthropic like the other import data. The category names may contain personal information if users enter such information as a category. If the setting "Use regional terms" is on (default), the two-letter country code of the device region (e.g. "AT") is also transmitted and passed on to Anthropic like the other import data; it is used only to apply regional terms when translating. An automatically detected cover/preview image is downloaded directly from the source and stored locally (and via iCloud); it can be replaced or removed in the recipe editor. The image host receives only the recipe origin (scheme, host, optional port) as Referer, never its path, query, fragment or credentials. No cookies or stored credentials are sent, including on redirects. If the restricted request fails, the optional photo is omitted. For the origin of these images and related third-party rights, see the Legal Notice.
Services involved:
- Own server (Cloudflare Workers): relay service, fetches the target page or API for web-link and YouTube imports and forwards content to the Anthropic Claude API. The server logs request metadata as an operational side effect (see section 5). Infrastructure operator: Cloudflare, Inc., USA - see section 6.
- Anthropic Claude API: processes photo content, PDF page images or extracted text (including from PDFs) to generate title, ingredients, preparation steps, nutrition estimate, and translation/unit conversion where needed. Anthropic's terms of service and privacy policy for API customers apply. Anthropic, PBC, USA - see section 6.
- YouTube Data API (Google): transmits the video ID to retrieve title/description/thumbnail of the public video and, where available, a comment by the channel owner. YouTube's terms of service and Google's privacy policy also apply. No video or audio is downloaded or stored. Google LLC, USA - see section 6.
- TikTok (fetched directly by the app): the public video page and the public, key-free oEmbed endpoint, called on your device, see below. TikTok's terms of service apply.
- Instagram/Pinterest (Meta Platforms Inc. / Pinterest Inc.; fetched directly by the app): the public post page, fetched on your device to extract title/caption and a preview image, see below.
Direct retrieval from video platforms: for links to TikTok, Instagram and Pinterest, the app retrieves the publicly visible description and the preview image address from the platform itself, not through our server. Your device connects to the platform; it receives your IP address, the time, the requested link and the usual technical details of a web request, and processes them as an independent controller under its own privacy policy. The app sends no cookies, no credentials, no household identifier and no subscription or device attestation to the platform. We do not receive the platform's response; only afterwards does our server receive the text that was read (not the link) to create the recipe. You can avoid this by not starting the import; if reading fails, you can paste the description yourself or enter the recipe by hand. The legal basis is the import function you requested (performance of a contract under the GDPR).
Transmission is encrypted (HTTPS). Storage in the app only after confirming the result in the editor. An import can be cancelled at any time before it is submitted; a request already sent to the server or the services listed above cannot be recalled afterwards.
Note on photos and PDFs: a photographed recipe or a recipe provided as a PDF (e.g. a handwritten card) may contain names, handwriting, or notes belonging to third parties alongside the recipe itself. Please only use your own or otherwise unproblematic material.
Note on result quality: AI-generated or translated information (in particular ingredients, allergens, quantities, and cooking temperatures) is an automated estimate. Please double-check before cooking, especially in case of allergies or intolerances.
Legal basis
Processing for the AI features serves to fulfil the actively requested recipe import (Art. 6(1)(b) GDPR). The app shows a notice naming the services involved before the first use of an AI feature; afterwards, use can be stopped at any time by simply not using the feature. The rest of the app's functionality remains fully usable independently of this.
5. Server-side quota and subscription checks
To prevent abuse of the paid AI calls and to correctly unlock a subscription, the app sends with every AI request a household identifier (a UUID generated once per household and synced via iCloud, unrelated to any name or Apple ID), for an active subscription the original transaction ID of the purchase, and, for free-tier use, an Apple App Attest device proof (see below).
The server stores, keyed to this household identifier:
- Total number of free imports used (purpose: enforcing the one-time free quota) - since this quota is one-time rather than monthly, this value is kept indefinitely, until deleted on request (see section 12) - it is not automatically deleted for inactivity
- Number of subscription imports and AI costs per calendar month, for monthly quota and operational budget checks; kept until deletion on request, with no automatic inactivity expiry
- Original transaction IDs bound to the household, binding/verification and expiry timestamps, an exclusive entitlement owner record, transfer timestamps and a deletion index; used for subscription verification and abuse prevention, kept until deletion on request
- Last-seen IP address (purpose: abuse detection), roughly 40 days, overwritten on every use and then automatically deleted
- Per device: a device identifier plus a public key and a signature counter (Apple App Attest, purpose: protecting the free quota against maliciously invented household identifiers without introducing a user account) - kept indefinitely, until deleted on request (see section 12) - not automatically deleted for inactivity
This data is visible per household in an access-protected operational dashboard, without names or Apple IDs. The original transaction ID is stored as a server-side binding and sent to Apple's App Store Server API. A truncated portion may appear in rare error/abuse logs.
Product analytics exports usage and cost values per household and month as CSV, without household IDs, IP addresses or replacement identifiers. Months with fewer than five included households are suppressed; rows are randomly reordered on each export. There is no identifier for linking months. Exact individual values may nevertheless be matched to the operational dashboard, so complete anonymity is not promised. Necessary household counters and the dashboard serve quota, cost and abuse checks separately.
Apple App Attest (device proof for the free quota): before a household without a subscription uses an AI feature, Apple's operating-system service App Attest confirms that the request comes from a genuine, unmodified installation of this app on genuine Apple hardware - without requiring an account or password. The device generates its own device-bound cryptographic key and has it attested once by Apple's attestation service; the app communicates directly with Apple's servers for this. The server stores up to 10 such device identifiers per household (public key + signature counter, no real names or device models) and rejects any free-tier request that does not carry a valid, matching proof. An active subscription does not need this extra check - the purchase itself is already a stronger proof.
Legal basis: legitimate interest (Art. 6(1)(f) GDPR) in cost control and abuse prevention; subscription access serves contract performance (Art. 6(1)(b)). Producing identifier-free product statistics is separately based on the interest in improving the app. Objections solely to that additional processing are assessed individually and, where upheld, implemented by excluding the household before further analysis, without automatically disabling AI imports. Objections to necessary quota/abuse checks are assessed separately under Art. 21. In the event of an objection or erasure, existing CSV copies are also reviewed; if a row cannot be reliably identified, the affected export file is deleted.
Deletion: email support@culinoria.com with the household identifier (Settings → Legal). Household counters, device identifiers, IP, binding index and ownership data for entitlements still owned by that household are removed. IP data also expires after roughly 40 days. Older downloaded identifiable CSV copies and third-party logs are outside this endpoint and must be checked separately in the request process.
Cloudflare Workers Logs
The server runs on Cloudflare Workers; Cloudflare logs request metadata (including IP address and requested path) as an operational side effect via Cloudflare Workers Logs. These logs serve error diagnosis and service operation and are subject to Cloudflare's own retention periods.
6. Transfers to third countries (USA)
Cloudflare, Anthropic, and Google (YouTube Data API) are based in, or operate servers in, the USA. The following safeguards under Art. 44 et seq. GDPR apply to transfers of personal data to them:
- Cloudflare, Inc. is certified under the EU-US Data Privacy Framework. The transfer relies on the European Commission's adequacy decision of 10 July 2023 (Art. 45 GDPR). In addition, the data processing agreement with Cloudflare includes the EU Standard Contractual Clauses.
- Anthropic is not certified under the Data Privacy Framework. The transfer relies on the EU Standard Contractual Clauses (Implementing Decision (EU) 2021/914), which form part of the data processing agreement with Anthropic (Art. 46(2)(c) GDPR).
- Google LLC is certified under the EU-US Data Privacy Framework (Art. 45 GDPR). Our server sends only the video identifier to the YouTube Data API.
Last verified: 10 October 2026. We will send a copy of the Standard Contractual Clauses on request to support@culinoria.com.
7. Share household
A household's entire data set - recipes (incl. photos), ratings, cooking history, meal plan, and shopping list - can be shared with one specific invited person via Apple's CloudKit sharing feature (CKShare, called "Share household"/"Invite person" in the app, under Settings).
- Sharing happens exclusively through a direct invitation to a specific Apple ID (no forwardable public link) - the invited person can get read or write access to the entire household, depending on the chosen permission.
- Once accepted, all listed data syncs bidirectionally via Apple's iCloud infrastructure.
- The provider never has access to shared content; Apple's privacy policy for iCloud applies.
- Shares can be revoked at any time via the system share dialog.
The subscription is shared through this app household. A subscriber unlocks it for the household's members; all share 200 AI imports per calendar month. Apple Family Sharing of the subscription is not offered.
8. Subscription and purchase handling
There are 5 free AI imports in total per household, or 200 AI imports per calendar month with a subscription (photo, PDF, web link, YouTube, social and a description pasted by you; an import with several pages or photos counts as one import). Shopping-list merging and sorting run locally and are free for everyone.
- Purchase handling exclusively via Apple (StoreKit/App Store Connect); no payment data reaches the provider.
- Subscription status is determined locally via StoreKit at purchase time. With every AI request, the subscription's original transaction ID is sent to the provider's own server and from there to Apple's App Store Server API for verification (see section 5) - so subscription status does NOT stay purely on-device.
- The subscription applies to the app household and its shared quota. At least one valid bound entitlement unlocks all members. Two subscriptions do not increase the quota. On a household change, the binding follows the buyer once their device registers the new household with the server; the previous household loses that entitlement.
- Cancellation/refunds exclusively via Apple ID settings or Apple's refund process.
9. Advertising and tracking
No advertising, no analytics/tracking SDK, no data sharing for advertising purposes.
10. Retention periods at a glance
- Local and iCloud data: until deleted in the app, the app is uninstalled, or a share is revoked.
- Anthropic API: according to Anthropic, API inputs and outputs are generally retained for up to 30 days and then deleted. Exceptions (e.g. abuse review, legal obligations) may apply. Details: https://privacy.claude.com/en/articles/7996866-how-long-do-you-store-my-organization-s-data
- Requests to Google (by our server) and TikTok/Instagram/Pinterest (by your device): per their own retention policies; no further server-side storage of the recipe content itself.
- Last-seen IP address: see section 5, roughly 40 days, deleted automatically.
- Household identifier, subscription bindings/ownership data and usage/cost counters: see section 5, no automatic deletion period (deleted on request).
- App Attest device identifiers (device proof for the free quota): see section 5, no automatic deletion period (deleted on request, together with the usage/cost counters).
- Cloudflare Workers Logs: see section 5, per Cloudflare's own retention periods.
11. iCloud backup
The app container may be part of iOS's native iCloud backup; this is governed by Apple's privacy policy and the user's iCloud settings.
12. Rights under the GDPR
Access (Art. 15), rectification (Art. 16), erasure (Art. 17 - for local/iCloud data, independently via the app or uninstallation; for server-side data see section 5), restriction of processing (Art. 18), data portability (Art. 20), objection (Art. 21, in particular against the processing described in section 5 that relies on Art. 6(1)(f)), complaint to a supervisory authority (Art. 77) - the competent authority is:
Österreichische Datenschutzbehörde (Austrian Data Protection Authority) Barichgasse 40-42 1030 Vienna Austria
Data subjects may also lodge a complaint with the supervisory authority of their habitual residence, place of work or place of the alleged infringement (Art. 77 GDPR).
Contact for inquiries to the data controller: see section 1.
13. Changes
Updated whenever actual data processing changes. Current version available at the URL registered in App Store Connect.