03 · Your content
We take no licence.
Most messengers ask you to grant a broad, worldwide, royalty-free licence to your content, and then explain that it is only for running the service. Livara asks for the narrow version: permission to store your content and transmit it to the people you send it to, and to synchronise it to your own devices. Nothing about training, promotion, or sublicensing to anyone else.
For direct and private-group text and media this is close to academic — the server holds ciphertext it cannot open, so the permission covers moving an opaque blob from one device to another. It matters more for the parts that are server-readable in 0.1.68 (69): channel content, profiles, stories and account metadata. Those are listed as readable on the security page rather than left for you to infer.
You are responsible for what you send. If you do not have the right to share a file, sending it through an encrypted channel does not give you one.
Moderation, honestly
Livara can act on reports about content it can read, and can suspend an account on other evidence — abuse of the API, mass account creation, credible reports of harm. It cannot moderate what it cannot read, and it will not add a way to read it. That is the cost of the guarantee on the front page, stated here so it is never a surprise later.
Livara AI
In the web app and Livara App, AI is included by default when you accept the account rules during signup. This one-time consent covers all your current and future chats for answers, reminders, summaries and related AI features. Existing accounts are asked once. Relevant message text, including private messages decrypted on your device, may be sent to Livara's configured AI service. You can turn AI off in Settings at any time. Text completion is requested with Ctrl+Space. This consent does not enable model training.
The standard Android app retains its own AI controls and consent flow. AI output can be wrong; review answers before relying on them.