Skip to content

Your data in Europe, and only commitments we can prove

Where the data lives, what we guarantee, the diagnostic's exception explained, and our commitments when we work on your tools.

Updated October 9, 20264 min read

Security is often presented through broad promises. For a small company handing its e-mails, documents and invoices to a provider, those promises are hard to check.

We hold ourselves to a simple rule: only present as guaranteed what we can show, in the code or in the hosting, and say plainly what is our way of working. Here is what that looks like, without dressing anything up, including the exception we make for the diagnostic on this site.

Where the data goes

When we build a company's central brain (a single place linked to all its tools, which knows what they know and answers with its sources) or a piece of custom software, the data has to live somewhere. Here is exactly where:

  • The data of the services we run is hosted by Supabase in Frankfurt, in the European Union.
  • Pages are computed by Vercel's servers in Frankfurt. Public files, such as images, are served by Vercel's global network. Vercel Inc. is an American company.
  • The automatic e-mails of our services go through Brevo, in the European Union.
  • For AI, our rule in production, meaning the services actually in use: only AI models called in the European Union, at AWS Bedrock, in its European regions.

So we do not say "everything is in Europe": public files go through a global network, and the diagnostic on this site is an exception, as explained below.

What we guarantee today

Each point on this list matches something that exists in the code or the hosting, and that we can show you:

  • Each organization, meaning a company's own space, is isolated by the database itself. Automated tests check this isolation on every change to the code.
  • Two-factor authentication is mandatory for owners and administrators.
  • Sign-in attempts are limited, per account and per network address.
  • Sign-ins, exports and changes of permissions are recorded in an activity log.
  • Strict security headers protect the pages, and session cookies only apply to the address of the service concerned, not to the whole domain.

Isolation deserves a closer look. It is not enforced by the screen, but by the database itself. Even a request made with a person's account, outside the screen, is refused in the same way. That is what makes it possible to check it with tests on every change, rather than claim it once and for all.

The diagnostic's exception, explained plainly

The diagnostic offered on this site is the one exception to our AI rule. To write the scenarios of your report, we send your answers to the form to Anthropic's API, that is, its AI service, in the United States. Only your answers to the questions, never the contact details asked for at the end (first name, company and e-mail address); simply avoid writing personal information in the free answers. Anthropic keeps them for up to 30 days, longer for content flagged under its usage policies.

The score, the map of your tools and the estimate of what the situation costs you do not go through AI: they are calculated by fixed rules. And if the AI is unavailable, the report still comes out, with standard scenarios taken from a library. As for how long things are kept:

  • Partial answers, without an e-mail address, are erased after 30 days.
  • Complete diagnostics are kept for 24 months.
  • The link to your report works for 90 days.
  • Diagnostic data lives in separate tables that only the server can reach.

Your answers and your contact details also reach our mailbox, at Google Workspace, which tells us about each new diagnostic. If you book a call, iClosed's calendar receives what you enter to book. The subprocessors registry says what each one processes.

Our commitments when we work on your tools

When you entrust us with your tools, further rules apply. They hold for every client, and the data processing agreement we will sign with you will set them out.

  • A data processing agreement is signed before any access.
  • We start by reading: our first access is read-only.
  • Every action is prepared, then approved by you. Nothing is sent, paid or deleted without your approval, and no money moves.
  • What is read in a tool, an e-mail or a document, stays data, never an instruction: it cannot trigger any action.
  • We remove our access at the end of the engagement.
  • You own your data.

Why we only show what we can prove

Displaying a commitment you cannot prove is asking for trust you have not earned yet. We would rather offer a shorter list that is true, and that you can check. If something is not listed here, it is because we cannot prove it yet.

We only present as guaranteed what the code and the hosting guarantee.

If you want to know what we would connect in your company, and how, the online diagnostic takes about four minutes and gives you a personalized report. You now know exactly what happens to your answers.

Read next

  • Practical AI

    Prepared by AI, approved by you

    Follow-ups, drafts, reconciliations, dashboards: what AI can prepare in a small business, and why you are always the one who approves.

    4 min read

  • Method

    Why connect your tools before replacing them

    Your information is scattered across your tools. Connecting them first, without changing anything, lets you find it all with one question, with sources.

    4 min read

From reading to your own case

The diagnostic draws the map of your tools and offers you three scenarios written for you.