Skip to main content

​
Definition

The prompt injection test (built with Llama) checks for prompt injection, malicious strings and jailbreak attempts in the input data of your system.

​
Taxonomy

  • Task types: LLM.
  • Availability: and .

​
Why it matters

  • Prompt injection is a type of attack that exploits an AI system and deviates it from its intended behavior.
  • It is important to detect and prevent prompt injection attacks to ensure the reliability and security of your system.

​
Judge explanations

You can add LLM judge explanations to rows that the test flags:
  1. In the create or edit test flow, open Advanced settings.
  2. Under LLM as a judge, turn on Enable LLM as a judge explanations.
  3. Use Judge model to keep the project’s default LLM-as-a-judge configuration or select a different model for this test.
The Judge explanation results column shows why the row was flagged, where the detection appears, and the relevant excerpt. The judge can identify a likely false positive, but its explanation is advisory and does not change the test count or verdict. Prompt-injection results also support the Summarize drill-down when the run has per-row explanations.

​
Test configuration examples

If you are writing a tests.json, here are valid configurations for the prompt injection test. Set explain_with_llm to true to request explanations. You can also provide llm_evaluator to override the project’s LLM-as-a-judge settings for this explanation pass.