Glossary

DEFINITION

Data Minimization

The privacy principle of collecting and processing only the minimum personal data necessary for a specific, stated purpose.

In depth

Data minimization is a foundational principle of modern privacy law. GDPR Article 5(1)(c) requires that personal data be 'adequate, relevant and limited to what is necessary in relation to the purposes for which they are processed.' CCPA's related concept of 'reasonable expectation' constrains secondary uses. The principle applies with force to AI interactions: sending a full contract to a model when only a specific clause needs analysis, or pasting an entire patient record when only the diagnosis is relevant, violates data minimization even if the AI provider has a compliant data processing agreement. AI systems make minimization harder because the natural tendency is to provide maximum context for better output — but every additional identifier in a prompt expands the blast radius if the data is retained, logged, or extracted by a prompt injection attack.

Examples

  • Sharing only the relevant clause of a contract with an AI redlining tool rather than the full agreement with all parties named.
  • Sending only the diagnosis code and treatment category to an AI summarisation tool rather than the full clinical note with patient demographics.
  • Asking an AI to analyse aggregate sales trends rather than providing individual customer transaction records.

How Locke handles data minimization

Locke enforces data minimization at the moment of AI interaction. By detecting unnecessary identifiers in a prompt — names, account numbers, dates of birth — and offering to remove or mask them before submission, Locke operationalises the minimization principle at the point where it matters most: just before the data leaves your device.

See the product

Related terms