PII & DATA LOSS PREVENTION
Protect sensitive data in every prompt
Automatically detect and redact PII, API keys, and sensitive information before it reaches your LLM. Reversible tokenization keeps context intact.
What the DLP path does
- 43 PII Entity Types Detected
- Comprehensive coverage across 11+ countries including SSNs, credit cards, IBANs, passport numbers, tax IDs, national health numbers, driving licenses, and more, with checksum validation for structured identifiers.
- Reversible Tokenization
- Replace each value with a reversible placeholder so the model keeps reference consistency, then restore the real value in the response.
- Secret Key Detection
- Automatically detect and block API keys, tokens, and secrets from being sent to LLMs using entropy analysis and provider-specific pattern matching.
- Custom Patterns
- Add your own regex patterns to detect domain-specific sensitive data like internal IDs or codes.
- Configurable Actions
- Choose to redact, mask, tokenize reversibly, or block requests containing sensitive information.
- Compliance Ready
- GDPR, CCPA, HIPAA-ready controls with audit logs and data handling documentation.
01
02
03
04
05
06
How PII protection works
- 01
Detect
ML models and pattern matching identify 43 entity types of PII across 11+ countries in incoming prompts.
- 02
Redact or Replace
Sensitive data is redacted or replaced with reversible placeholders before reaching the LLM.
- 03
Log & Comply
All PII detections are logged for compliance. Original data never leaves your control.
Automatic PII protection
python
# User input with sensitive data
user_input = """
Contact me at john.doe@company.com or 555-123-4567.
My SSN is 123-45-6789 and credit card is 4111-1111-1111-1111.
"""
# PromptGuard automatically protects
response = client.chat.completions.create(
model="gpt-5-nano",
messages=[{"role": "user", "content": user_input}]
)
# What the LLM sees:
# "Contact me at [EMAIL_REDACTED] or [PHONE_REDACTED].
# My SSN is [SSN_REDACTED] and credit card is [CARD_REDACTED]."
# Or with mode="tokenize", reversibly:
# "Contact me at PG:PII:EMAIL:0 or PG:PII:PHONE_US:0..."
# — and the real values are restored in the response.Why PromptGuard DLP?
- 43 PII entity types across 11+ countries
- Reversible tokenization preserves context
- Zero-config protection
- GDPR, CCPA, HIPAA ready
- Included in all plans
Protect sensitive data
Stop PII from leaking to LLMs. Automatic detection and protection in every request.