---
aliases:
  en:
  - alignment auditing
  - alignment audits
concept_id: alignment_audit
preferred:
  en: alignment audit
relations:
  relations:
    broader:
    - evaluation
    prerequisite:
    - ai_alignment
    related:
    - interpretability
    - red_teaming
status: active
---

# alignment audit

A systematic investigation of an AI model or system intended to uncover hidden or unintended alignment-relevant objectives, propensities, or behaviors, potentially using behavioral testing, training-data analysis, interpretability methods, and other investigative tools.

## broader

- [evaluation](../terminology/evaluation.txt)

## prerequisite

- [AI alignment](../terminology/ai_alignment.txt)

## related

- [interpretability](../terminology/interpretability.txt)
- [red teaming](../terminology/red_teaming.txt)

[Corpus index](../index.txt)