PII
Data Pribadi
Information that identifies a person, directly or through a combination of details.
PII is any information that identifies a person: a name and phone number, obviously, but also an email address, an employee id, a customer number, a device id, or a combination of ordinary details that happens to be unique.
It matters for AI because a model does not need to be malicious to leak it. A prompt that includes a record, a log that records the prompt, a model fine-tuned on customer records: all three spread the data further than intended, and all three are ordinary engineering decisions rather than attacks.
The minimum useful practice is a redaction step before anything personal goes into a prompt, and a rule that customer data never goes to a consumer model endpoint. Both are cheap. Discovering a leak in production is not.
Related terms
Data Privacy
The rules about whose personal data you may hold, and what you must do with it.
Data Governance
The rules and habits that decide what data exists, who may use it, and how long it lives.
Guardrails
Checks that stop or constrain an action before it happens.
Local LLM
Running the model on your own machine or server, with no provider in the loop.
Compliance
Meeting the legal rules that apply to what you built, in the place you built it.
Want this working inside your team?
It starts with one conversation about your workflow.