A service’s price, revenue model and collected data are three separate questions; each answer needs evidence.
A messenger being free does not by itself prove that it sells users’ data. Services may be funded through donations, subscriptions, business services or advertising. The right questions are what data the service collects, why it needs it, how long it keeps it and with whom it shares it.
End-to-end encryption is only part of this assessment. The encrypted content of a message and information about the timing and route of communication are not the same issue.
Why is “if it’s free, you’re the product” not enough?
This slogan may draw attention to economic incentives, but it is not a substitute for evidence. For example, Signal’s donation page explains its nonprofit model supported by user donations. We cannot infer that it sells the communication graph merely because it is free.
Conversely, paying does not guarantee data minimization. A paid service may retain data for operations, support or product analytics. Assess price, company ownership, privacy policy and technical architecture separately.
| Funding model | Useful question to investigate |
|---|---|
| Donations | How transparent are the main funding sources and operating costs? |
| User subscriptions | What data does payment generate, and which providers are involved? |
| Business services | How are personal-account and business data kept separate? |
| Advertising | What data are ads based on, and is it connected to private conversations? |
The existence of a revenue model does not, on its own, establish all of a service’s data practices.
What is metadata, and why is it sensitive?
Metadata is information surrounding communication: timing, approximate data size, delivery routes or group members, depending on the service’s design. Combining these data may reveal communication patterns even when message text is not read.
However, not every connection allows identity, conversation topic or precise location to be inferred with certainty. Data quality and context matter. Recording an IP address, for example, is not the same as having an exact GPS location.
Nor should the value of message content be dismissed. The sensitivity of metadata does not mean that the contents of a contract, photograph or private conversation have “little value.”
How should we read a privacy policy?
Instead of looking for a marketing phrase, record the following:
- Data type: Identifiers, phone numbers, contacts, IP addresses, timestamps, notification tokens, error reports or payment data?
- Purpose: Routing, abuse prevention, synchronization, analytics or advertising? Do not confuse a broad, vague purpose with a specific explanation.
- Recipient: The service itself, a hosting contractor, notification services, a payment provider or an advertising partner?
- Retention: What is the deletion period or criterion? How does account deletion affect backups and security records?
- Control: What is optional, and which feature does not work without it?
If the answer is unclear, write “unknown.” No evidence of sales is not proof of no collection, and collection does not automatically prove sales.
What does Livara explicitly state?
Livara’s security page says that operational metadata is retained for delivery and synchronization, including peer relationships, group membership, timing, ciphertext size and delivery status. The same page states that LSS1 is not yet connected to the normal message-sending path.
Livara therefore cannot be described as “completely anonymous,” nor its database as “containing no useful information.” Encrypted private content and observable metadata are two different parts. For the rules governing data use, read the privacy policy alongside the technical architecture, not instead of it.
SRP authentication does not determine the revenue model or metadata collection either. This protocol concerns login; using it does not make selling data mathematically impossible. Post-quantum encryption is likewise not synonymous with hiding the communication graph.
Build a useful comparison
For two or three options, create a table with columns for “required data,” “optional data,” “service access,” “retention period,” “revenue model” and “source and date.” Do not force an unknown feature into a green or red checkmark. If you use an enterprise service, also add the contract, device management and the organization’s retention requirements.
In summary: Privacy claims need support from evidence and design. Being free is not grounds for condemnation, and neither payment nor the name of an algorithm establishes immunity.
