AI privacy practice library32
DiscoverMAP-01
AI data-processing inventory
Record personal and linkable data collected, generated, inferred, transformed, retained, disclosed, shared, used, and disposed across each AI use case.
DesignMAP-02
End-to-end data lineage
Trace data from source and permission through transformation, feature use, model or agent processing, outputs, decisions, storage, sharing, correction, and deletion.
OperateMAP-03
Derived and inferred data map
Identify predictions, embeddings, classifications, scores, summaries, prompts, logs, and other outputs that reveal or create information about individuals.
ReassessMAP-04
Sensitive and high-impact data classification
Classify financial, health, biometric, location, behavioral, protected-class, household, claims, and other sensitive or consequential data.
DiscoverPUR-01
Specific purpose definition
State the concrete insurance purpose, affected people, expected benefit, decisions supported, users, boundaries, and evidence of necessity.
DesignPUR-02
Authority and permission basis
Document applicable legal, contractual, regulatory, consent, authorization, and policy bases for every material data action.
OperatePUR-03
Compatible-use and reuse review
Assess whether secondary use, enrichment, model training, evaluation, sharing, or new inference is compatible with the original context and commitments.
ReassessPUR-04
Prohibited and restricted use controls
Prevent disallowed collection, inference, combination, targeting, surveillance, decision use, disclosure, training, or cross-context transfer.
DiscoverNOT-01
Layered AI privacy notice
Explain material data sources, purposes, inferences, recipients, retention, decision uses, rights, choices, assistance, and changes in accessible layers.
DesignNOT-02
Consent and authorization design
Use specific, informed, accessible, unbundled, revocable consent or authorization when required or selected as the appropriate control.
OperateNOT-03
Preference and choice management
Capture and enforce consumer choices consistently across channels, products, systems, models, vendors, and future processing.
ReassessNOT-04
Notice and consent evidence
Retain the notice version, timing, channel, audience, language, accessibility, choice presented, response, withdrawal, and downstream enforcement.
DiscoverMIN-01
Collection and feature minimization
Limit raw data, prompts, context, features, outputs, logs, and telemetry to what is proportionate and needed for the approved purpose.
DesignMIN-02
Data and feature necessity review
Challenge whether each data element or proxy materially contributes to the purpose and whether a less intrusive alternative is practical.
OperateMIN-03
Retention and disposal schedule
Apply event-based retention to raw data, features, prompts, outputs, embeddings, logs, caches, backups, and vendor copies, with legal holds handled.
ReassessMIN-04
De-identification and re-identification control
Validate de-identification claims, linkage risk, access restrictions, release context, contractual controls, and periodic re-identification testing.
DiscoverRGT-01
Request intake and identity assurance
Offer accessible request channels and use proportionate identity verification without collecting excessive additional information.
DesignRGT-02
Access and explanation fulfillment
Locate and provide covered data, sources, recipients, categories, purposes, significant inferences, decision context, and required explanations.
OperateRGT-03
Correction deletion and restriction
Evaluate and fulfill applicable correction, deletion, restriction, objection, and withdrawal requests with documented exceptions.
ReassessRGT-04
Downstream propagation and closure
Propagate approved changes to systems, features, vendors, models where applicable, future decisions, and records; confirm completion to the requester.
DiscoverENG-01
AI privacy impact assessment
Assess data actions, affected people, context, likelihood, privacy consequences, mitigations, residual risk, approval, and reassessment triggers.
DesignENG-02
Privacy requirements by design
Translate privacy objectives into architecture, interfaces, defaults, permissions, separation, logging, retention, review, and acceptance criteria.
OperateENG-03
Leakage memorization and inference testing
Test prompts, outputs, retrieval, models, agents, logs, embeddings, and interfaces for unintended disclosure, reconstruction, or sensitive inference.
ReassessENG-04
Privacy-enhancing technique validation
Validate tokenization, masking, encryption, differential privacy, synthetic data, confidential computing, and other techniques against the intended threat and use.
DiscoverTPR-01
Vendor purpose and use restriction
Limit provider access, use, disclosure, sale, enrichment, model improvement, monitoring, and support activity to documented approved purposes.
DesignTPR-02
Subprocessor and data-transfer control
Identify subprocessors, locations, transfer mechanisms, access paths, changes, flow-down terms, and oversight evidence.
OperateTPR-03
Provider retention and training restriction
Control prompt, file, output, telemetry, support, and backup retention and prohibit unauthorized provider or foundation-model training.
ReassessTPR-04
Exit return and deletion assurance
Plan portability, transition, access removal, copy return, deletion, backup expiration, certificate evidence, and residual dependency at exit.
DiscoverGOV-01
Privacy accountability and escalation
Define owners, decision rights, risk acceptance, privacy-office authority, escalation, board reporting, and coordination with security, legal, compliance, and AI governance.
DesignGOV-02
Access logging and investigation
Record and review human, application, model, agent, vendor, and administrative access and enable timely investigation of anomalous use.
OperateGOV-03
Privacy metrics and control monitoring
Measure inventory, minimization, consent, requests, retention, vendors, testing, incidents, exceptions, and privacy outcomes with thresholds and action.
ReassessGOV-04
Privacy incident and breach coordination
Connect privacy-event assessment with security response, legal analysis, consumer harm mitigation, notification, correction, regulator response, and lessons learned.