Authenticated account access
Protected application routes verify the signed-in account before serving private data.
To unify conversations, provide search, and answer questions with AI, Claire synchronizes message data through its own service. Connected networks continue to apply their own security and encryption model. When an AI feature runs, selected conversation context may be sent to the configured AI provider.
These are implementation-backed product controls, not a substitute for a completed independent security assessment.
Protected application routes verify the signed-in account before serving private data.
The API uses security headers, a production origin allowlist, and separate rate limits for login and AI routes.
AI is an explicit feature surface. We will describe the active mode and provider rather than presenting AI as invisible background processing.
Private desktop-only mode remains in development until storage, egress, search, and AI behavior are independently verified.
We prefer showing the flow to making blanket privacy claims.
A message originates in WhatsApp, Telegram, Instagram, or another connected service, under that service’s own security model.
Claire receives and normalizes the data needed to show the unified inbox, preserve history, power search, and support connected features.
Ask Claire, summaries, and suggestions use selected conversation context with the configured AI mode. This is a separate boundary.
We will only upgrade our language when a real configuration, test suite, and review demonstrate the behavior in production.
mautrix supports end-to-bridge encryption, but it must be enabled and verified for Claire’s bridge deployment before we market it as a protection.
We need audited outbound-network controls, local search and media behavior, local model rules, and documented mobile boundaries before making a local-only claim.
We will publish a practical security review, incident-contact path, retention/deletion controls, and a configuration-specific data-flow statement.