It means deciding what the AI needs to know about your people – and what it does not.
One prompt is enough for personal data to leave a company. So the question is not ‘cloud: yes or no’, but: what exactly goes out, and in what form?
Usable
SecureData preserves format, matching, linking and grouping.
Understandable
Pseudonymisation replaces recognised identities with placeholders – the model understands the sentence.
Mapping in Switzerland
The mapping of the placeholders is held by Praevana in Switzerland.
One prompt is enough.
As soon as data flows into an AI system – as a prompt, as context, as reference material – it leaves your sphere of control. A single prompt with sensitive content can end up with the model provider, appear in logs or be output again by indirect routes.
The solution.
Two protection concepts, one consistent approach: Prewen cleanly separates two tasks – because they need different tools.
01
SecureData: working with the value without knowing it
Structured identifiers are protected in a format-preserving way: consistent, linkable and reversible for authorised roles. Typical places: databases, SaaS fields, test data, analytics, backups, logs.
02
Pseudonymisation with Praevana: understanding the sentence without knowing the identity
In free text sent to a language model, Praevana replaces recognised identities with placeholders such as <PERSON_1> or <CH_IBAN_1> and reinserts them in the response. The placeholders apply across the entire conversation.
Usability and understandability.
SecureData protects the value and preserves its usability: format, matching, linking, grouping. Pseudonymisation protects the identity and leaves everything else in plaintext – which is why anything that does not name a person anyway remains computable there.
Where protection applies.
Key and mapping sovereignty.
SecureData derives the key for each field from the identity instead of storing it: no key is stored per protected value. If we operate SecureData for you, the environment is located in Switzerland.
Why Prewen.
Praevana
Prewen AG's platform for handling sensitive data securely in AI projects.
OpenText SecureData
Format-preserving encryption for fields and data streams.
prewen.ai and KI-ATELIER
From assessing the potential to secure implementation – introducing AI with an eye on data control.
Prewention
Set up AI projects with data kept secure.
Together with you, we clarify which data your AI project needs and how it stays protected.