---
aliases:
  en:
  - AI red teaming
  - red-teaming
  - red team testing
concept_id: red_teaming
preferred:
  en: red teaming
relations:
  relations:
    broader:
    - evaluation
status: active
---

# red teaming

A structured adversarial testing process in which testers actively probe an AI system, product, or safety mechanism to discover harmful capabilities, unsafe behaviors, vulnerabilities, or failure modes by adopting the perspective and methods of a motivated adversary.

## broader

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

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