How does masked.chat actually protect my privacy?
Before your message ever leaves your device, masked.chat detects and replaces all personally identifiable information (PII) — names, emails, phone numbers, addresses, and more — with neutral placeholders. This masking happens locally on your device, so neither the AI model you're talking to nor masked.chat's own servers can see your real data. Your identity stays in an encrypted vault that never leaves your control.
What technology detects my PII?
We use a combination of context-aware rule-based detection and a lightweight local LLM running on your device. This hybrid approach catches both structured data (like phone numbers and emails) and more nuanced contextual PII (like a name embedded naturally in a sentence). Because detection runs locally, your data is never sent anywhere just to be analyzed.
What is the identity vault?
The identity vault is a secure, encrypted store that holds the mapping between your real PII and the masked placeholders. It allows AI responses to be "unmasked" back into meaningful answers on your side — for example, replacing a placeholder like [PERSON_1] with the actual name you typed. The vault is local by default. If you create an account, an encrypted copy is synced to our servers solely to enable cross-device continuity — we cannot decrypt it.
Does masked.chat work with any AI model?
Yes. masked.chat sits as a privacy layer between you and the underlying AI model, so it is model-agnostic. Whether you are using a major provider or another supported model, the masking process applies the same way. The AI receives only the anonymized version of your input and never has access to your real identity.