Glossary
PII Redaction
Also known as: data redaction, PII scrubbing
Definition
PII redaction is the process of detecting personal identifying information in text and removing, masking, or replacing it before that text reaches storage, a third-party AI provider, or downstream processing. Categories typically include names, SSNs, card numbers, emails, addresses, dates of birth, and any identifier that could singly or in combination identify a person.
Why it matters
Whether PII redaction happens at ingestion or after storage is the difference between a routine compliance posture and a HIPAA breach assessment. Redacting at ingestion means sensitive content never lands in the trace database, the embedding store, or the third-party AI provider's logs. Redacting after the fact means it did, and the team is now in the cleanup-and-disclosure business.
GDPR Article 25 (data protection by design) and HIPAA's de-identification standards both expect redaction to be a designed property of the pipeline, not a remedial control. PCI DSS requires it for cardholder data. State privacy laws (CCPA, CPA, VCDPA) add their own variants.
In practice
Prism Guardrails redact PII before any storage step. The detector catalog covers 30+ identifier categories with validators (Luhn for cards, ABA for routing, IBAN checksum, VIN) so false positives stay low on niche identifiers. Prism X applies the same detection at the browser, blocking PII from ever leaving an employee's machine into ChatGPT, Claude, Gemini, or Copilot.
Related
More glossary terms
Start tracing in 5 minutes
One SDK. Five minutes. Full audit trails, PII redaction, and guardrail enforcement, from day one.