Demo limitations
What the interactive demo does and does not show.
The demo on /demo/ is a concept illustration built on synthetic data. This page states its boundaries explicitly so that nothing here is mistaken for a live product capability.
What the demo actually does
- Loads three preset scenarios with clearly synthetic identifiers such as
SYNTHETIC_TENANT_001. - Applies a small, rule-based redaction pass in the browser so you can see substituted values.
- Shows which fields would be classified as sensitive and which token would replace them.
- Lets you choose one of three disclosure outcomes: approved cloud, local-only, or blocked and review-required.
- Shows a pre-written simulated output and a simulated restoration step.
What the demo does not do
- It does not call the Claude API. No network request is made by the demo.
- It does not upload, store, or transmit any file or text you type.
- It does not read any real database, tenancy, or document store.
- It does not write a token map to
localStorage, cookies, or session storage. - It does not run the PrivaGate AI detection engine. The pattern matching shown is simplified and is not production Japanese PII detection.
- It does not use artificial latency to imitate model inference. Transitions are immediate.
- It does not accept real contracts, real tenant data, or real client information.
Why the demo is drawn this narrowly
A marketing website should never become an unmanaged side channel for confidential documents. Keeping the demo browser-only, synthetic, and deterministic means a visitor can understand the intended data flow without any real information leaving their machine, and without the website needing to hold any sensitive data at all.
Where the demo output comes from
The structured output shown after a policy decision is a static, hand-written illustration stored in the plugin fixtures. It is labelled simulated in the interface. It is not a model response and is not presented as one.