How Sannvit Uses AI

Sannvit, Inc. · Effective August 12, 2026 · Version 1.0

Sannvit builds evidence infrastructure for AI systems — our product exists so that organizations can prove what their AI did. That obliges us to hold our own AI use to the standard we sell. This statement describes how we actually operate, under our internal AI Usage Policy, adopted by our board on August 10, 2026. It is written to be checked, not admired.

Three rules we do not break

1. No AI output containing unverified facts, citations, or claims leaves this company. Every external document, message, or publication gets named-human verification before it ships.
2. No confidential information — ours or anyone else's — goes into AI tools that can train on it. We use AI only through channels contractually barred from training on our inputs and outputs, and some categories (credentials, personal tax data, certain unfiled intellectual property) don't enter any AI tool at all.
3. No AI acts externally on its own. Nothing is sent, filed, signed, published, or transacted by an AI system without a named human approving that specific action first. We sell sealed human-oversight evidence; we run on it.

These are not aspirations; they are prohibitions in a board-adopted policy that no internal approval can waive.

When you interact with our AI

If you interact with a generative AI system operated by Sannvit, it will tell you so — our product demo agent identifies itself as AI at the start of every demonstration, whether or not you ask. Where Utah's Artificial Intelligence Policy Act requires disclosure, we comply; where our own standard is stricter than the statute, we follow our standard. And if you ask whether a document we sent you was drafted with AI assistance, we will answer plainly.

How decisions about AI are made here

Every AI use falls in one of three tiers: routine internal work is approved by the responsible founder; anything customer-facing, any new AI tool, and any AI in outbound sales requires both founders; anything touching the rules above requires both founders and counsel. New tools are reviewed for their training-on-data terms before adoption — a vendor right to train on our confidential data is an automatic no. AI incidents are reported same-day, recorded, and never minimized; the policy is reviewed at least every six months.

What our product does — and doesn't — do with AI

The Sannvit product observes and seals records of your AI systems' activity. The product itself is deterministic evidence infrastructure: it does not make automated decisions about people, and its detection logic is deliberately rule-based and explainable. It is self-hosted in your infrastructure — your AI activity records never pass through Sannvit's systems.

Questions

Customer due-diligence teams are welcome to send AI-use questionnaires; we answer them from this policy and its internal counterpart, completely. Contact: brad@sannvit.com · Sannvit, Inc., 176 N 170 E, Orem, UT 84057.