Personally Identifiable Information (PII) is any data that can identify a specific person: names, addresses, phone numbers, email addresses, government ID numbers, and similar details. PII Redaction lets a corpus mask this information automatically, so assistants can use the useful parts of a document without exposing anyone's personal data.Why PII Redaction matters#
Many valuable documents also contain personal details. Support tickets include customer names and emails, HR files include employee records, and contracts include signatories' contact information. Without redaction, an assistant could repeat those details to anyone who asks. PII Redaction lets you put this content to work while protecting the people in it.Privacy by default: Personal data is masked before assistants ever see it.
Safer sharing: You can connect a corpus to several assistants, or share it widely, with less risk of leaking personal information.
Compliance support: Redaction helps you meet data protection obligations and your organization's privacy policies.
Trust: Customers and employees can rely on your assistants not to reveal their details.
How it works#
1.
You turn on PII Redaction for the corpus.
2.
Documents are scanned when uploaded. Personal information in each document is detected automatically.
3.
PII is masked. Detected details are hidden or replaced with placeholders.
4.
Assistants use the redacted version. When an assistant searches the corpus, it works with the masked content, so answers stay useful without exposing personal data.
With PII Redaction turned on, only TXT, DOCX, and PDF files can be uploaded to the corpus.
Use cases#
Customer support content: Turn on PII Redaction before uploading ticket exports or chat transcripts, so a support assistant can learn from past cases without repeating customers' names and emails.HR and recruiting: Redact employee records, interview notes, and résumés before an HR assistant uses them.Widely shared corpora: For a corpus shared across the whole organization, where you can't control who asks the assistant questions, redaction keeps personal details out of every answer.To hide personal information in an assistant's conversations instead of its documents, use the PII Redaction guardrail. See Understanding Guardrails.Where to turn it on#
When you create a corpus: turn on Enable PII Redaction in the Create Corpus panel. See Assistant Corpus Setup. For an existing corpus: open the corpus Settings tab and use Enable PII Redaction under PII Settings. See Assistant Corpus Settings. 