WHIZZVOICE

Legal

AI Disclosure

Effective 13 July 2026 · Version 1.0

Whizz Voice powers AI voice agents that talk to people on the phone and the web. The people on those calls deserve a plain statement of what the AI does, how it tells them it is an AI, what it is prevented from doing, and what can be asked of the companies that deploy it. This page is that statement, written for callers, customers, and regulators alike.

1. What the AI does, on a call

  • Recognize. A speech-to-text model transcribes the caller as they speak, streaming and diarized — Arabic dialects as spoken, digits and terms code-switched to English the way callers actually talk. This is acoustic transcription within one call; no voiceprint is stored or matched across calls.
  • Reason.A frontier language model reads the live transcript against the customer's persona, knowledge base, and rules, and decides what to say and which tool to call. Answers are grounded in the customer's content through retrieval, not invented.
  • Speak. A dialect-native text-to-speech model synthesizes the reply in the selected voice and dialect and streams it back — typically first audio in about half a second. The voice is a synthetic persona, not a clone of a specific real person.
  • Act.The agent calls the customer's functions and webhooks mid-call — look up a record, book a slot, send a message, hand off to a human — with only the data needed to complete the task.

2. Telling callers it's an AI

Whether an agent discloses that it is an automated system is a configuration gate in Whizz Voice, on by default. Customers deploying agents where disclosure is required — many jurisdictions require it — must keep it on, and our Terms prohibit disabling it to deceive callers. When disclosure is on, the caller is told, in their own language, that they are speaking with an AI.

3. Recording and consent

Recording is off unless the customer turns it on. When it is on, a recording-consent gate can require the agent to announce the recording and capture consent before anything is stored, with the moment logged against the call. Customers may also run agents in zero-retention mode, where audio and transcripts are processed in-memory and not persisted. The lawful basis for recording sits with the customer.

4. Human oversight

An agent does what the customer configured it to do and nothing more. It can hand off to a human with full context when it hits its limits or the caller asks. Whizz Voicedoes not make employment, credit, legal, or medical decisions; customers are responsible for not deploying an agent for decisions it isn't fit to make without human oversight.

5. Who is affected, and what they can ask for

  • Callers who speak with an agent have data rights (access, deletion, objection) exercised through the company that deployed the agent, per our Privacy Policy. Requests sent to support@whizztech.ai are routed to the responsible organization.
  • Anyonecan ask the company they called whether the call was recorded and request the transcript or its deletion, subject to that company's policies.

6. Regulatory posture, in plain language

  • AI-disclosure and calling law — requirements to disclose automated systems, to honor do-not-call and calling-time rules, and to obtain consent to call sit with the customer deploying the agent. Whizz Voice provides the disclosure and consent gates; our Terms require customers to use them lawfully.
  • UAE PDPL & TDRA and GDPR — we act as processor for call content, under a data-processing agreement, with the controller obligations resting on the customer. Consent and outreach defaults are built with UAE and Saudi PDPL and UAE TDRA expectations in mind.
  • EU AI Act — customers deploying voice agents for EU callers should assess their obligations, including transparency duties toward people interacting with an AI system. This is not legal advice; obligations depend on how and where you deploy.

7. What we do not do

  • We do not deploy agents that hide being an AI where disclosure is required — the gate is on by default and our Terms forbid disabling it to deceive.
  • We do not clone a specific real person's voice or identity; the voices are synthetic personas.
  • We do not identify callers biometrically and store no voiceprints across calls.
  • We do not train models on call audio, transcripts, or outcomes.
  • We do not claim our own SOC 2 or HIPAA certification — we build on processors that hold them.

8. Questions

If your compliance team needs specifics — the models used per stage, data flows, subprocessors, or the disclosure and consent gates — write to support@whizztech.ai. Engineers answer this inbox.