Security and trust
You should know exactly what happens to your information.
Trust is earned with specifics. This page says how Averyniq is designed to handle your information, and where the specifics are set out in writing for each pilot before any data is shared.
What you will be told
Seven things, before any data is shared
What you provide
The files, connections or access you choose to give, and nothing beyond what the agreed scope needs.
What is processed
What is read, extracted or analysed, and which features use an AI provider.
Where processing happens
The locations involved, stated only where they have been verified.
Who may receive it
The AI providers and subprocessors that may receive information, named for your pilot.
What is kept, and for how long
Retention and deletion terms, agreed in writing.
What you control
Which sources are connected, who can see what, and how you revoke access.
What happens when an answer cannot be supported
The platform is designed to say that it cannot substantiate an answer, rather than guess.
Working with your data
Five ways of working, by arrangement
Averyniq is designed around five ways for a customer to bring data to the platform. Which of them a given pilot uses is agreed in writing for that pilot, and every one is designed to keep least privilege, tenant isolation, your consent, an audit trail and the ability to revoke.
Secure upload
You upload files and recordings. The platform is designed to read them within bounded limits and to keep a record of what was read and what was not.
Governed read-only connection
You connect a database or application with a read-only account that you create. Read-only is designed to be enforced by the account's own permissions and checked on the connection, never by an instruction to a model.
Customer-hosted connector
A small connector runs inside your network and only makes outbound requests, so you open no inbound path. Designed for organisations that cannot expose a database.
Isolated or dedicated deployment
For organisations that require their own environment. Scope and terms are agreed individually.
Natural-language questions over your authorised sources
You ask in plain language; the platform is designed to answer only from sources you have authorised, to cite them, and to say so when it cannot.
AI answers
How answers are meant to be earned
Grounded in your sources
An answer is designed to cite the records it rests on, and only records you are allowed to see.
Fact kept apart from inference
Statements are designed to be labelled as sourced fact, calculation, interpretation, possibility or recommendation, or as unknown.
Numbers from code
Calculations are designed to be done by deterministic code with stated units and rounding, not by a language model.
A person decides
Consequential steps are designed to need a person's approval, and nothing is sent, filed or settled for you.
No claim of perfection
No AI system never errs. Averyniq is designed to measure its own quality against independent tests and to withhold an answer it cannot support.
What this page does not claim
No certification is held or claimed. This page makes no statement about where data is processed, how long it is kept, whether a provider uses information to train models, or that deletion is permanent. Those are disclosed, with the evidence for each, in the written terms of a pilot, before any data is shared.