What is PII?
Personally identifiable information (PII) is any data that can identify a specific person, alone or combined with other data, such as a name, email address, phone number or ID number.
In more detail
Definitions vary by law. The GDPR uses 'personal data': any information relating to an identified or identifiable person. In AI projects, PII turns up in places you do not expect: invoices with sole-trader names, support transcripts, email signatures, call recordings. Handle it by collecting less, redacting before text reaches a model or a log, limiting who can see it and deleting on schedule. Also decide where the model runs, since sending PII to a third-party API creates a data transfer you need to document. A common pattern is a redaction step that swaps names for tokens before the LLM call and restores them afterward.
Example
A support copilot strips card numbers and national ID numbers from tickets before drafting a reply. The original stays in the helpdesk, and the model sees only masked text.