Privacy
AI receives text with your permission. Backup, sync and exports are separate choices that can copy originals off your phone.
AI reads text with your permission
Your phone reads PDF text or recognises text in scanned pages locally. With your explicit permission, text and layout information go through our Cloudflare Worker to Workers AI. The text may contain personal data. Pictures and PDFs are not sent to AI. Our Worker neither logs nor stores document text or model responses.
Your library, backups and exports
The pilot includes a local document library and project finances. You can choose to save originals and reviewed details, group them into projects and track budgets, invoices and payments you record yourself. Reopening needs no new AI reading. Unsaved reviews disappear when the session ends.
Device backup and private iCloud sync are separate, optional choices, off by default. Enabling them can copy originals, reviewed details, projects and payments off your phone. They do not pass through the AI service. iCloud sync is built but has not been verified between real devices; system restore and the production schema also await verification.
For exports you choose the destination, which may be a cloud service. A portable backup also contains originals and personal data and is not password-protected by the app. Portable backup and iCloud sync have a 250 MB limit. You manage exported copies. Turning off sync does not delete the cloud copy; cloud deletion is a separate choice.
A Swedish sole trader’s registration number is a personal identity number. It may be in the library you choose to save and its optional backup/sync copies, but is not stored in our server registers, statistics or logs. We do not build a company register from documents.
Operational records
The AI service keeps pseudonymous records for quotas and budgets: a random pilot ID, request IDs and usage measures, cleared after their month plus at most one more. Separate failure reports contain fixed error codes and limited measures, never document content; they expire after 14 days and are removed at the next daily cleanup, normally within 15 days.
Optional quality measures
Quality sharing requires a separate optional choice. Only a closed set of quality measures is accepted, with no identities, exact prices, document dates, pictures or free text. Contributions can be withdrawn with their receipt within 30 days; afterwards only aggregate counters remain. These measures are not used for model training or contractor rankings.
This website
This is an information website. We do not accept documents, payments or new interest registrations here. The website has no analytics scripts or advertising pixels of its own. Ordinary connection metadata is processed by the hosting infrastructure.
Earlier waiting list
If you previously signed up, a record may exist with four fields: email (email address), country (chosen country), locale (the page’s language/market code) and created_at (time of first signup). The new information website writes no such records. Existing records have not been deleted by this website change. Email hello@readmyinvoice.com for questions and deletion requests. Do not send invoices or other documents by email.
Data location and providers
The waiting list’s D1 database was created with the WEUR location hint. This guarantees neither a particular location nor EU jurisdiction. We do not promise EU-only processing or zero retention at the AI provider. Cloudflare states that customer content is not used for model training without consent. Provider operations, caching and agreements need review before launch.
Before public launch
The responsible seller/data controller, agreements, transfers, legal basis, retention and App Store disclosures still need completion before public launch. Apple/RevenueCat are the planned route for consumer purchases; payments are disabled and documents must not be sent there as customer attributes.