Safety policies applied to all gateway traffic — pre-request and post-response
2 Active
Guardrails are applied at the gateway level
All traffic through WhiteGator passes through active guardrail policies before reaching the LLM and after receiving the response. Configure per-project policies in Organizations → Project Settings.
🔒
PII Masking
MASK
PII_MASKING
Automatically detects and masks Personally Identifiable Information such as emails, phone numbers, SSNs, and credit card numbers.
Stage: PRE_REQUEST + POST_RESPONSEScope: All Projects
Enforced
🛡️
Prompt Injection Shield
BLOCK
PROMPT_INJECTION
Detects and blocks prompt injection attacks that attempt to override system instructions or extract sensitive data.
Stage: PRE_REQUEST + POST_RESPONSEScope: All Projects
Enforced
⚠️
Toxicity Filter
BLOCKDisabled
TOXICITY
Filters out hate speech, harassment, adult content, and other toxic language from both inputs and outputs.
Stage: PRE_REQUEST + POST_RESPONSEScope: All Projects
Off
🔍
Custom Regex Rules
LOG ONLYDisabled
REGEX_BLOCK
Custom regular expression rules to block or redact specific patterns in request or response content.
Stage: PRE_REQUEST + POST_RESPONSEScope: All Projects
Off
Guardrail Pipeline
1
Request arrives at gateway
Client sends to /v1/chat/completions
2
PRE_REQUEST guardrails fire
PII masking, injection detection applied
3
Request sent to provider
Sanitized payload forwarded to LLM
4
POST_RESPONSE guardrails fire
Response scanned before returning to client
5
Log & audit
Guardrail trigger recorded in audit log
Action Types
BLOCK
Reject the request entirely and return a 400 error. Best for security-critical rules.
MASK
Replace sensitive content with a placeholder (e.g. [REDACTED]). Request still proceeds.
LOG ONLY
Allow the request but record a guardrail trigger in the audit log for review.