Your voice is sensitive. We treat it that way.

Voiceport may process writing samples, uploaded documents, customer-submitted URLs, public social captions, transcript exports, phone conversation transcripts, and the generated voice.md artifact.

We treat that material as customer-owned source data. The product should collect only what it needs, send only the context required for generation or voice transport, and avoid retaining raw material longer than the customer-facing retention promise allows.

The Starter product can be short-retention: create the Voice Profile, deliver it, support the order, then delete or redact raw source material on the published schedule.

The goal is not to ask you to believe Voiceport is impossible to compromise. The goal is to avoid keeping more sensitive raw material than the product needs.

Some future upgrades, such as refreshes, team profiles, agent-ready exports, and quarterly updates, work better if Voiceport keeps a profile history. That retention should be a visible customer choice, not a hidden default. You should be able to use fresh material instead.

Voiceport uses specialized providers for storage, AI generation, calls, payments, email, and analytics. Those providers should receive the minimum data needed for their job. Payment tools do not need your writing. Analytics tools do not need your voice.md. Call transport does not need unrelated account history.

Before broad customer testing, provider accounts must be Voiceport-owned, isolated from unrelated businesses, and reviewed for model-training, retention, logging, deletion, access-control, and encryption behavior.