Skip to main content

COMPEL Glossary / red-team-for-llms

Red-team (for LLMs)

A structured adversarial exercise against an LLM feature using human, automated, or hybrid techniques drawn from MITRE ATLAS or OWASP LLM Top 10 to discover failure modes before attackers do.

What this means in practice

Extends classical red-teaming with LLM-specific techniques such as jailbreak suite probes, indirect prompt injection payloads, and tool-use attack chains.

Synonyms

LLM red-teaming , adversarial LLM testing , generative red team

See also

  • Evaluation harness — The infrastructure that runs capability, regression, safety, and human-review evaluations on an LLM feature on a defined cadence.
  • Jailbreak — A user-crafted prompt pattern that bypasses a model's safety training to elicit restricted behavior.
  • Guardrail — A control placed between the user or environment and an LLM that blocks, rewrites, or classifies content at one of four architectural layers: input filter, policy filter, output filter, or tool-call validator.