The concepts we work with.

Recognizable details such as names or companies are replaced with placeholders before text is processed further. Whoever is authorized can translate the placeholders back.

Pseudonymized data is still personal data under the GDPR. Only what is recognized gets replaced, and the context can still allow conclusions.

Only what a specific purpose needs is processed and stored – and not kept longer than necessary. Less data means less that can go wrong.

It reduces risk but does not replace security: whatever is needed is still processed.

Usage is measured only where it helps improve a product – anonymously, in aggregate and without cookies.

Even anonymous counts are data about use. What a product measures is set out in its privacy policy.

Data is stored on servers in the EU, so EU law applies to its storage.

This covers where data is stored. Third-party services, such as AI models, may work outside the EU – which ones is listed in the privacy policy.

Parts of a system run separately from each other, each with only the rights it needs. A fault in one part does not automatically open up the others.

It limits the consequences of a fault; it does not prevent the fault itself.

Credentials for connected services, such as access tokens, are stored encrypted. Someone who can only read the database cannot use them.

To be used, credentials have to be decrypted. If the running system itself is compromised, encryption at rest does not help.