# AI Testing Standards · AIT 

## Standard Metadata

- Area: Quality
- Owner: Technical Leadership group
- Last reviewed: 2026-09-09
- Next review due from: 2026-12-01
- Enforcement: Code review
- Status: Active

## Applications that use large language models must conduct LLM evaluations to test model outputs · AIT-01 · **MUST** · **DEV/TEST**

Audacia's preferred approach to LLM evaluation is captured as a [sensible default](https://dev.azure.com/audacia/Audacia/_wiki/wikis/Audacia.wiki/16232/LLM-Testing).