The privacy layer for AI · Zürich, Switzerland
Clausurus.
Your words,
on your terms.
Ask any question. See what personal details were detected, how they were replaced, and exactly what your chosen AI model receives.
No login. Chat history stays in this browser, not a profile.
Start a conversation
See what the model sees.
Choose an example or write your own. You’ll review the filtered message before it goes to the chat model.
Or try an example
No login needed. Chat history stays in this browser. Sent messages in original and filtered form are privately saved for debugging and usage; deleting local history doesn’t erase those records. Optional deep scanning sends your original text to Swiss-hosted Apertus before filtering.
01 — The idea
A checkpoint before the model.
Most AI tools ask you to trust what happens after you press send. Clausurus puts a visible privacy gateway in between. It looks for direct identifiers and contextual clues, replaces them with consistent placeholders, and lets you inspect the outgoing message before you approve it.
Explore the approach02 — The flow
Privacy you can actually see.
Find the sensitive parts
Names, emails, card numbers and subtle clues such as “the only pharmacist in town” can be flagged before a request goes out.
Replace, not erase
Identifiers become readable placeholders like [NAME_1], keeping a conversation useful without handing over the original detail.
Inspect the handoff
Compare your original with the filtered version in the side panel. Turn on approval to decide whether it should be sent.
03 — Principles
Built for scrutiny.
Visible data flow
The same inspect-before-send principle guides the planned API: organizations should be able to trace what was detected, transformed and forwarded.
Swiss-rooted, honest about boundaries
The project is based in Zürich and uses a Swiss-hosted Apertus service for optional deep scanning. The chosen chat model may be hosted outside Switzerland; the app shows its data handling status, so Swiss-only processing is not guaranteed for every model.
Open at the foundation
The basic version is open source, so the approach can be examined and built upon.
For teams
Bring the gateway into your pipeline.
Interested in building privacy-aware AI into your company's workflow? Let's talk about your requirements and what an integration could look like.
Contact us