PII Plugin

The PII (Personally Identifiable Information) plugin tests an AI system's ability to protect sensitive personal data. It helps ensure AI responses safeguard user privacy and comply with data protection principles.

Repeating personal information supplied by the user in the same conversation is not a PII leak. Strategy grading and regrading saved results use the recorded attack input, including per-turn transforms. Crescendo, Custom, and GOAT also include prior target conversation when available to distinguish user-provided information from newly disclosed data. Hydra and Goblin grade the current turn without adding prior conversation history.

Provider-reported chat prompts retain prior user/assistant context without treating system or tool messages as user-supplied information. A grade from an earlier turn is not reused for a different returned prompt, response, or assertion configuration; results without a verifiable strategy grade are graded again.

Configuration

Include the PII plugin in your LLM red teaming setup:

redteam:
  plugins:
    - pii

Or specify individual subcategories:

redteam:
  plugins:
    - pii:direct
    - pii:session
    - pii:social
    - pii:api-db

You can also provide a specific name to target:

redteam:
  plugins:
    - id: pii:direct
      config:
        name: 'Jane Doe'

Subcategories

Subcategory	Plugin ID	Description
Direct Disclosure	pii:direct	Tests direct revelation of PII when asked
Session Leakage	pii:session	Checks PII leaks across user sessions
Social Engineering	pii:social	Evaluates manipulation resistance
API/Database Access	pii:api-db	Ensures no unauthorized PII retrieval attempts