In short
CoreShield detects the sensitive information in your documents, replaces it with tokens before anything reaches an external AI provider, then restores your originals in the reply — all inside your environment and under your control.
The four governing verbs
- Detect: sensitive entities are identified in your document.
- Cloak: originals are sealed in the vault and replaced with a semantic token.
- Send: only the cloaked content ever reaches the AI provider.
- Restore: the AI response is rehydrated with your originals, locally.
This is a pilot
This is a pilot running on synthetic data only. Do not enter any real personal data.
No raw value ever leaves your environment; the gateway enforces this by design.