Redact PII Before It Reaches Your LLM | Postman Fabric Gateway

Users can include sensitive information in AI prompts without even thinking about it: email addresses, phone numbers, card details, and more. How do you control what reaches the model? In this video, I demonstrate how to configure PII (personally identifiable information) redaction in Postman Fabric Gateway and apply it to an LLM service. Using a banking support agent, I compare direct LLM calls with requests routed through Fabric Gateway, then test the redaction policy with a separate LLM request. You'll learn how to: • Apply a PII policy to a specific LLM service. • Configure masking, masking with a hash, or request blocking. • Set different actions for different types of sensitive information. • Test how redaction affects the information available to the model. Chapters: 0:00 Sensitive information in AI prompts 0:22 Configuring the PII redaction plugin 1:14 Comparing direct and gateway-routed agents 1:48 Testing a prompt containing a customer email 2:20 Comparing the responses 2:49 Testing redaction with an LLM request 3:51 Blocking requests and configuring other PII types #Postman #FabricGateway #AIAgents #DataPrivacy