Legal

AI transparency statement

Version 1.1 · 5 August 2026 · English only for now

loquaro is a voice AI agent that makes and answers phone calls for business clients. This statement says plainly how it works and which rules it always follows — the same rules our client contracts commit to. It is published by Developers Alliance LLC and versioned: when our practices change, this document changes with them.

What the agent is

The agent runs on Google’s Gemini real-time voice models, on the paid tier — which means conversations are not used to train models. Each client’s deployment runs on a server dedicated to that client, in a region chosen by the client’s policy. You can hear the agent yourself at loquaro.com/talk.

When it says it’s an AI

Whether the agent introduces itself as an AI in its greeting is decided per client, during pilot setup — markets and scripts differ. A disclosing greeting sounds like:

“Hello, this is the AI assistant calling on behalf of [company] about your payment due on Friday.”

Whatever the greeting says, the agent is built and instructed not to present itself as a person: asked directly whether it is an AI, it says that it is. This is not a client setting. It is also what the terms of the model provider we run on require of us, which is why we do not offer it as one.

What it may and may not say

• Every deployment runs on a script the client approved word by word. The agent speaks freely around that script — never past it. No invented facts, no promises or consequences the client didn’t authorize.
• If speech is unclear, it asks again rather than guessing — amounts, dates, and names are confirmed, not assumed.
• Disputes, hardship cases, and anyone who asks for a person are not argued with: the agent closes politely and escalates to the client’s team with the recording and transcript attached.

What it can’t do

• It cannot take payments during a call; it points to the payment channels the client already has, exactly as the script describes them.
• It does not give legal or financial advice, and it does not decide disputes — those go to humans.
• Like any speech system it can mishear, especially on poor lines — which is why unclear input is re-asked and every call is recorded and transcribed, so no outcome rests on what the agent thought it heard.
• The browser demo is a demo: a public test line with time limits, not a production deployment on a client’s script.

Who decides who gets called

The client decides which numbers are called, when, and on what script. The client is responsible for having the right to call those numbers under the law of its own market — consent, permitted calling hours, and any notice a recording requires — and confirms that in the client agreement. We do not source phone numbers and we do not decide who is contacted. What we do is configure the greeting and any recording notice to what the client approves, and keep the approved script on file.

What happens to call data

Calls are recorded, transcribed, and returned to the client as structured outcomes — all of it, or only what the client’s policy allows. Data is kept on the client’s schedule and deleted on it; demo conversations are not stored at all. Details, including the vendors involved, are in the privacy policy.

Status of this statement

These practices are self-attested — we have not yet undergone third-party certification — and contractually committed to clients. If the agent ever behaves outside these rules, we want to know: [email protected].