Glossary
Glossary: pseudonymization, de-identification and AI-ready copies
Each definition says what the term means, what it does not mean, and which published standard it comes from.
What is pseudonymization?
Pseudonymization replaces the details that identify a person, such as a name or ID number, with stand-ins like [PERSON_001], and keeps the link between stand-in and real value in a separate, protected key. Whoever holds the key can reverse it; whoever does not sees consistent placeholders. Because it is reversible, regulators such as the EDPB and the UK ICO treat pseudonymized data as still personal data.
What is de-identification?
De-identification is the general process of removing or transforming the details in a record that identify a person: removing direct identifiers such as names and ID numbers, and reducing quasi-identifiers such as dates, locations and occupations. It is an umbrella term. Pseudonymization is one form; anonymization is the irreversible end of it. A de-identified record can still be re-identified if enough detail remains.
What is a quasi-identifier?
A quasi-identifier, or indirect identifier, is a detail that does not name a person on its own but can identify them in combination with other details or outside knowledge. Ontario's privacy commissioner lists examples including date of birth or age, event dates, postal codes, building names, regions and profession. The classic case: ZIP code, birth date and sex together made most Americans unique in 1990 census data.
What is a re-identification key?
A re-identification key is the table, or secret, that maps each stand-in in a pseudonymized document back to the real value: [PERSON_001] to a client's name, [SIN_001] to their SIN. Whoever holds it can reverse the pseudonymization. NIST calls it a lookup table and says it "must be highly protected"; the EDPB says lookup tables are themselves personal data. Keep it away from the AI platform.
What is an approved copy?
An approved copy is PiBye's term for the version of a document that is ready to share with an AI: identifying details replaced with stand-ins, every replacement reviewed by a person, and the whole copy explicitly approved. It is pseudonymized, not anonymized. The original and the key that reverses the stand-ins stay on the Mac; only the approved copy is copied, saved or offered to an assistant.
What is a local MCP connector?
A local MCP connector is a program on your own computer that a desktop AI app, such as Claude Desktop or the ChatGPT desktop app, starts as a local process and talks to over standard input and output (STDIO) using the Model Context Protocol. The app can then call the tools it offers. The connector runs locally, but whatever a tool returns becomes part of the conversation the AI provider processes.