Skip to main content

COMPEL Glossary / ai-incident-for-llms

AI incident (for LLMs)

A subtype of the NIST AI RMF MANAGE 1.4 incident concept specific to LLM systems: confident-but-wrong answer, safety bypass, prompt injection success, sensitive data leakage, or policy-violating tool-call execution.

What this means in practice

Classification drives response severity and regulator-notification thresholds.

Synonyms

LLM incident , generative AI incident

See also

  • Model and prompt registry — A versioned inventory of models, system prompts, retrieval sources, and guardrails deployed in production.
  • Serious incident (Art. 3(49)) — Under Regulation (EU) 2024/1689, an incident or malfunction of an AI system that directly or indirectly leads to the death of a person or serious harm to a person's health, to a serious and irreversible disruption of critical infrastructure, to infringement of fundamental-rights obligations, or to serious harm to property or the environment..
  • Post-market monitoring (Art. 72) — Under Regulation (EU) 2024/1689, the ongoing, documented collection, analysis, and corrective-action process that providers of high-risk AI systems must operate after the system is placed on the market.