Skip to content

Draft — for counsel review

This document is a draft. It has not been reviewed or approved by counsel, it is not in effect, and nothing here is legal advice or an offer capable of acceptance. It is published so you can see exactly what we intend to put in front of counsel — the banner comes off only when they clear it.

AI Technology Disclosure

Version: v1.0 · Referenced by the Vocapable Master Services Agreement §4.

This notice explains, in plain language, what Vocapable’s AI technology is and what it means to rely on it. It is part of your agreement with us.

What the technology is

Vocapable’s voice agents are powered by artificial intelligence: large language models that generate what the agent says, speech-recognition models that transcribe what people say, and speech-synthesis models that produce the agent’s voice. Some of these models are built and operated by us; others are provided by third parties.

These systems are probabilistic. They do not follow a fixed script the way a recording does — they generate responses in the moment. That is what makes them useful, and it is also why they can be wrong.

What can go wrong

You should expect, and plan for, all of the following — before, during, and after calls:

  • An agent may say something inaccurate, incomplete, or invented (“hallucination”), even about your own product, prices, or policies, and even when your knowledge sources are correct.
  • An agent may misunderstand the person it’s speaking with, mishear words, or respond to something that wasn’t said.
  • An agent may deviate from your configuration — your prompts, objectives, and guardrails reduce this but cannot eliminate it.
  • Transcripts, summaries, extracted data, and scores may contain errors.
  • Behavior may change over time as models are updated, including third-party models we do not control.
  • Calls may fail, drop, or degrade for technical or carrier reasons.
  • The same input can produce different outputs on different runs.

This is an experimental, rapidly evolving technology under ongoing development. We test extensively — every agent version passes an evaluation gate before it can be published — but testing reduces error rates; it does not make errors impossible.

What we ask of you (and what you agreed to)

  • Review outputs before relying on them. Do not treat anything an agent said, or any transcript or summary, as verified fact without checking it.
  • Decide where human review belongs in your workflow. You know your use case; we don’t.
  • Never use the Service as the sole basis for decisions with legal, financial, medical, employment, or similarly significant effects on a person.
  • Tell people they’re talking to an AI. Every call opens with a non-removable AI disclosure — that’s a floor we enforce, not a ceiling. Your own use case may require more under applicable law.
  • Configure carefully. Your prompts, knowledge packs, and guardrails are yours; the agent speaks on your behalf and at your direction.

What this means legally

As set out in the Master Services Agreement: the Service and its outputs are provided “as is”; we do not warrant their accuracy; you are responsible for evaluating their fitness for your use; we are not responsible for third-party models and providers; and, as between you and us, you bear the risk of adverse outcomes caused by output errors, subject to the MSA’s limitation-of-liability and indemnification provisions. If any part of this notice conflicts with the MSA, the MSA controls.