It sounds to me like you've nailed precisely what the Jitsi README is talking about. Full context from the README (https://github.com/jitsi/jitsi-meet/blob/mast

It sounds to me like you've nailed precisely what the Jitsi README is talking about. Full context from the README (https://github.com/jitsi/jitsi-meet/blob/master/README.md https://github.com/jitsi/jitsi-meet/blob/master/README.md) is this: > WebRTC does not (yet) provide a way of conducting multi-party conversations with end-to-end encryption. Unless you consistently compare DTLS fingerprints with your peers vocally, the same goes for one-to-one calls. As a result, your stream is encrypted on the network but decrypted on the machine that hosts the bridge when using Jitsi Meet. > The Jitsi Meet architecture allows you to deploy your own version, including all server components. In that case, your security guarantees will be roughly equivalent to a direct one-to-one WebRTC call. This is the uniqueness of Jitsi Meet in terms of security.