Open security model
Don't trust us.
Don't trust us.
Verify us.
Security claims are cheap. This page names every mechanism behind ours — and is honest about today's limits.
「本当に安全なの？」に答えるページです。各主張がどの仕組みに対応するかを明記し、現時点での限界も隠さず記します。
What we do — and don't
We do
- ✓Encrypt every message end-to-end with AES-256-GCM.
- ✓Keep keys on your device, gated by passkey.
- ✓Let you verify a contact's fingerprint yourself.
- ✓Seed key material with quantum entropy.
We don't
- ✕Ask for accounts, phone numbers, or email.
- ✕Store messages on any server.
- ✕Keep logs we could be forced to hand over.
- ✕Claim to hide metadata we can't yet hide (see below).
The mechanisms
Identity Lock
Passkey
Your identity key never leaves the device, and only unlocks with your biometric.
Honest limitIf the device itself is compromised, passkey protection fails. Device lock and keeping your OS up to date are prerequisites.
Implemented
Transport
WebRTC · Peer-to-Peer
Devices talk directly. The signaling server only introduces peers — it never sees the conversation.
Honest limitTURN relay may be required to traverse NAT. The payload is still encrypted during relay, but connection metadata (who connected when) cannot be fully hidden.
Implemented
Encryption
AES-256-GCM
Every message is sealed with authenticated encryption — confidentiality and tamper-detection in one.
Implemented
Randomness
Quantum Entropy
Key material is seeded with true randomness from the ANU Quantum Random Numbers API — not a software PRNG.
Honest limitQuantum entropy improves key generation quality — it does not guarantee privacy on its own. The foundation remains: no server storage, E2E encryption, passkey. Falls back to OS-provided CSPRNG when the quantum source is unavailable.
Implemented
Read it. Then decide.
Transparency is the feature. The rest is up to you.