CONCEPT Cited by 2 sources
Purpose limitation¶
Definition¶
Purpose limitation is the data-protection principle (codified e.g. in GDPR Art. 5(1)(b)) that personal data may be collected and processed only for specified, explicit purposes and not further processed in ways incompatible with them. As a systems problem it reduces to: can we prove a piece of data only flows where it is allowed?
Why it is an infrastructure problem at scale¶
- Point-checking does not scale.
ifstatements and dataset ACLs need exhaustive, continuous audits and push teams to physically split data so each asset has one purpose; Meta found this unworkable across millions of assets and dozens of systems. (Source: sources/2024-08-31-meta-enforces-purpose-limitation-via-privacy-aware-infrastructure) - Lineage alone discovers but doesn't enforce — see concepts/data-lineage.
- Runtime concepts/information-flow-control — annotate data with purposes, propagate the label through the computation, and check every sink — is the durable primitive (Meta's Policy Zones).
- Enforcement must sit where context exists. Meta's PrivacyLib evaluates purpose checks in-process inside data clients, where caller identity, endpoint and purpose are still known, rather than at the datastore. (Source: sources/2026-10-27-meta-privacylib-privacy-controls-at-the-data-access-boundary)
Rollout discipline¶
Both Meta posts converge on observe-before-enforce: Policy Zones zones run in logging mode before enforcement mode (2024); PrivacyLib gates new enforcing modules behind a runtime rollout percentage and simulates annotation changes against observed flows before release (2026).
Seen in¶
- sources/2024-08-31-meta-enforces-purpose-limitation-via-privacy-aware-infrastructure — Policy Zones / PAI; why point-checking and lineage-only fail.
- sources/2026-10-27-meta-privacylib-privacy-controls-at-the-data-access-boundary — enforced as a PrivacyLib module across hundreds of systems; extended to AI-agent tool calls.
Related¶
- concepts/information-flow-control
- concepts/contextual-integrity — a broader normative framing of appropriate flows.
- concepts/policy-as-data
- systems/meta-privacy-aware-infrastructure