Agentic AI Reference Architecture

Safety & Security · Software component

Heuristic Jailbreak Detector

Software componentSafety & SecuritySafety, Security & Governancearc:HeuristicJailbreakDetector

A jailbreak detector that flags attacks from perplexity drops, prompt-injection patterns (e.g., regular expressions) and known jailbreak preambles.

Responsibility. Screens input quickly for documented attack signatures.

Also known as: Injection pattern check

Variant of Jailbreak Detector abstract

When to choose. Choose for fast initial screening of all traffic; effective against documented attacks but not novel ones.

triggers; is target of alternativeTois invoked byspecializesClassifier Jailbreak Detector: triggers; is target of alternativeToClassifier Jailbreak Det…Input Rail: is invoked byInput RailJailbreak Detector: specializesJailbreak Detector
Direct neighbourhood (hover for relationship types)

Relationships

is invoked by dependency

triggers dynamic

alternative to variability

Classification

Patterns
Perplexity analysisRegex pattern blockingKnown-preamble matching

Sources

  1. Ch7.1B: T. Nguyen, "Nvidia NIM and Colang," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 7.1B. ISBN: 9798244538229.
  2. Ref7.03: NVIDIA, "Overview," NVIDIA NeMo Guardrails Library Developer Guide. Accessed: Sep. 27, 2026. [Online]. Available: https://docs.nvidia.com/nemo/guardrails/about-nemo-guardrails-library/overview
  3. Ref9.04: "Safety Guardrails Implementation for Agent Systems," unpublished reference note (04-Safety-Guardrails-Implementation.md), Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam supplementary materials, 2026. unpublished note