PII and PHI in AI features
When a visitor to your knowledge base asks AI Answers a question or sends a ticket that is then parsed by AI Agent Assist, the contents are automatically passed to Microsoft Azure and Open AI (our AI model providers) for processing. This might be undesirable or noncompliant if sensitive data is included in the query.
Sensitive data could be:
Personally identifiable information (PII), like name or address.
Protected health information (PHI), like age or medical records.
On request, we can enable automatic PII and/or PHI removal, so those details are never sent to our AI vendors for processing.
Removing PII and PHI from queries
When PII and/or PHI removal is active, each query for AI Answers and AI Agent Assist is scanned for sensitive data before it's sent to the AI model. If sensitive data is detected, it is replaced by placeholders.
For example, if a user types My name is John Smith, call me at 555-123-4567, the query sent to the language model will be: My name is <PHI_NAME>, call me at <PHI_PHONE_OR_FAX>.
AI features will generate a complete response as they would do normally, but they will never process PII/PHI.
PII removal works in all languages supported by AI Answers. PHI removal is only supported in English.