OpenPrompts
← Back to catalog
NVIDIAGuardrailsSafety & Moderation

Fiddler (NeMo Guardrail)

flow fiddler user safety """ Use Fiddler's SLMs to detect jailbreaks """ $is_jailbreak = await CallFiddlerSafetyUserAction() if $is_jailbreak if $conf

flow fiddler user safety
  """
  Use Fiddler's SLMs to detect jailbreaks
  """
  $is_jailbreak = await CallFiddlerSafetyUserAction()

  if $is_jailbreak
    if $config.enable_rails_exceptions
      send JailbreakDetectionRailException(message="Jailbreak attempt detected. The user's prompt was identified as an attempted jailbreak. Please ensure your prompt adheres to the guidelines.")
    else
      bot refuse to respond
    abort


flow fiddler bot safety
  """
  Use Fiddler's SLMs to detect jailbreaks
  """
  $is_jailbreak = await CallFiddlerSafetyBotAction()

  if $is_jailbreak
    if $config.enable_rails_exceptions
      send JailbreakDetectionRailException(message="Jailbreak attempt detected. The user's prompt was identified as an attempted jailbreak. Please ensure your prompt adheres to the guidelines.")
    else
      bot refuse to respond
    abort

flow fiddler bot faithfulness
  """
  Use Fiddler's SLMs to detect hallucinations
  """
  $is_faithful = await CallFiddlerFaithfulnessAction()

  if $is_faithful
    if $config.enable_rails_exceptions
      send FaithfulnessRailException(message="Hallucination Detected. The output was a hallucination.")
    else
      bot refuse to respond
    abort
Flagged for review by the automated scan
  • AST01Safety bypass / jailbreakMalicious Skills · High

This is a heuristic text match and may be a false positive (security and coding prompts often mention these terms). Review the content before use. How we scan

Provider
NVIDIA
Origin
Official
Type
Guardrails
License
Apache-2.0
Language
English
Added
2025-09-03
#guardrail#nemo#rails#colang