Security
Security model
mAIsk exists to reduce accidental leakage of sensitive information into AI prompts. The product is intentionally local-first and workflow-oriented.
What it does
Replaces sensitive values with deterministic placeholders before text is sent to an external AI system, and restores the original data locally afterward.
What it does not do
It does not magically anonymize meaning, prevent contextual inference, or secure a compromised machine. It reduces risk. It does not eliminate it.
Local-first promise
No cloud dependency is required for masking and restoration. That is central to the product’s positioning and should remain clearly documented.
Workflow trust
The system relies on placeholder integrity and deterministic behavior. If placeholders are corrupted or altered, restoration must fail safely.