Know what stays here, and what you choose to share.

Local storage is not a private account

Saved details are encrypted with AES-256-GCM, but the usable key is stored in the same browser profile. That enables passwordless restoration. It is not an authentication boundary: another person using that profile, a malicious extension or compromised application code could access the details.

The app can hold up to 30 saved document versions within a 12 MB plaintext limit. Storage is tied to the exact address and browser profile. Clearing site data, closing private browsing or organisational policy can remove it. An installed app does not add a separate security layer.

Optional restore files contain the originals

An optional .restore-file moves an extracted original and matching list to another browser or address. It is not encrypted and has no password. Save it only somewhere appropriate, keep it separate from your AI copy, and never attach it to an AI service.

Shared-computer mode can download without saving new matching details in the tab. It does not erase earlier saved work. Starting again also clears only the current tab, not the saved-document list.

Website requests are different from document uploads

Vercel hosts the application. Loading it creates ordinary website requests, which can expose an IP address and request metadata to the hosting provider. The document workspace has no analytics or advertising scripts, remote fonts or paid AI calls. Optional analytics on the public pages is explained below.

The app caches static code and assets for offline use, not your document downloads. After that cache is ready, preparation and restoration can work offline. The external AI websites still need internet and have their own account settings and data policies.

Replacement is not guaranteed anonymisation

A reversible label is a pseudonymisation-style step, not proof that nobody can identify the person. Other wording, dates, locations, small groups or information held elsewhere may still identify them. The ICO distinguishes pseudonymisation from effective anonymisation; using labels does not itself establish legal compliance.

Your organisation must decide whether the document, device and AI account are suitable. Use the tool to reduce what you share, not to bypass a restriction. We do not claim automatic detection of all identifiers, guaranteed anonymity, regulatory compliance or layout-preserving document redaction.

Optional visitor analytics

If you choose Allow analytics, Google Analytics measures visits to the seven public information pages. It never runs in the document workspace or practice guide. Saying No thanks does not change the app.

We send a fixed public page address and title, a random visitor reference and a visit reference. We exclude URL queries, fragments, referral addresses, filenames, document text, replacement lists and AI answers. Google also receives ordinary connection and browser information, such as your IP address. This is optional measurement, not a claim of anonymous browsing.

Google's code runs in a browser-isolated frame without access to the app's saved documents or page contents. Advertising features and automatic interaction tracking are off. Your choice and optional random reference are stored in this browser for up to 180 days. Google also uses analytics cookies on metrics.adaped.com for up to 180 days. Use Analytics choices at the bottom of any public page, then No thanks, to stop future collection. That does not erase data already received by Google. Event-level retention is two months; Google's separate user-level retention setting is 14 months. Activity does not reset the user-retention period.

Before you start