Privacy notice
Last updated: 19 August 2026
Oscura is currently published in demo mode. The service genuinely works, but the payment system is simulated: no payment details are collected and no amount is ever charged.
1. Who is responsible
The controller of this processing is Noelia Fabiana Conde, the individual who operates this service, reachable at sapienzaia2024@gmail.com.
No data protection officer has been appointed, because this processing does not fall into the cases where the GDPR requires one (Art. 37). For anything concerning personal data, the email address above is the point of contact.
To be completed when payments go live: the operator's VAT or tax number and registered address. Payment platforms ask for this before approving an account, and it belongs here and in the terms of service.
2. The part that matters most: your documents
The files you process are not uploaded to our servers and are not stored by us. The PDF is opened by your browser on your own computer. When a page is a scan, the optical recognition of its text also happens entirely in your browser.
The only thing that leaves your device is the text of each page, sent to our server and from there to the provider of the AI model, which tells us which fragments are personal data. The page image is never transmitted. The redacted PDF is rebuilt on your computer and downloaded straight from there.
The text you send is processed only for as long as it takes to answer the request: it is not written to disk, not indexed, and does not become part of any archive. Your account history keeps only the page number, the date and the cost in credits — never the content.
If you process documents containing other people's data (clients, employees, counterparties), you are the controller for that data and we act as your processor, limited to the moment of processing.
3. Use of artificial intelligence
This service runs on artificial intelligence, and we say so openly. The model used is Claude, provided by Anthropic PBC.
- What it does: it receives the text of a page and proposes which fragments are the personal data of natural persons. Nothing else.
- What it does not do: it never receives the image of your document, it does not decide on its own what gets covered, it produces no new text and it does not modify your document. The final PDF is built on your computer, not by the AI.
- The decision stays human. The AI's proposal is shown as a list you can switch on, switch off and add to by hand. Nothing is downloaded until you explicitly confirm that you have checked the document. This is therefore not automated decision-making within the meaning of Art. 22 GDPR: the final decision, and the responsibility for it, belong to whoever uses the service.
- Training: Anthropic does not use content submitted through its API to train its models. We do not use it to train anything and we do not keep it.
- Stated limits: an AI model can miss a piece of data or flag something that is not personal data. Poor scans and handwriting reduce reading reliability further. That is why the human check is not an extra — it is part of the procedure.
- Beyond the AI: some categories with a recognisable shape (emails, IBANs, tax codes, phone numbers, dates of birth) are also picked up by deterministic rules that run in your browser, independently of the model.
If you would rather not rely on artificial intelligence, you can still cover any area of the document by hand: the tool works without accepting the model's proposals.
4. What we keep
| Data | Why | How long |
|---|---|---|
| Email address | To identify your account and let you sign in | While the account exists |
| Hashed password | To verify sign-in. Not readable, not even by us | While the account exists |
| Credit balance and plan | To make the subscription work | While the account exists |
| History (date, page number, cost) | To show you what you have used | Last 60 operations |
| Approval record (date, file name, page and redaction counts) | To give you a trace of who approved which document and when, should you ever need to show it. Nothing of the content is kept | Last 60 operations |
| Session cookie | To keep you signed in | 30 days |
| Server technical logs | Security and fault diagnosis | Per the hosting provider's retention |
We do not ask for your name, address, phone number or payment details.
5. Legal basis
We process this data to perform the contract created when you open an account (Art. 6.1.b GDPR) and, for the technical security logs, on the basis of our legitimate interest in protecting the service (Art. 6.1.f GDPR).
6. Cookies
We use one cookie, oscura_sesion, strictly necessary to keep you signed in. It is not a
profiling cookie, it serves no advertising and it requires no consent. We use no analytics and no
third-party tracking. Your chosen language is also stored in your browser and never leaves your
device.
7. Who else sees the data
- Netlify — hosts the site, the account store and the server functions.
- Anthropic — provides the AI model that analyses the page text.
- Google Fonts — serves the site's typefaces.
We do not sell, rent or share your data with anyone else. Some of these providers process data outside the European Union, under the standard contractual clauses approved by the European Commission.
8. Your rights
You can ask at any time to access your data, correct it, delete it, restrict its processing, object to it, or receive it in a readable format. Write to sapienzaia2024@gmail.com and we answer within 30 days. You may also lodge a complaint with your national data protection authority.
To close your account and erase everything we hold about you, one email is enough.
9. Security
The site is served only over an encrypted connection. Passwords are stored with a one-way derivation function and cannot be recovered in the clear. The AI service keys stay on the server and are never reachable from the browser.
10. Changes
If this notice changes materially — for example when payments become real and a payment provider is added — we will update the date at the top and tell you by email.