Legal Register
Visual Infographic
I-25 AI & Analytics 👤 Chief Data Officers, AI Engineers & ML Architects

Derived-Data & AI Reach Map

Consent State Propagation across Feature Stores, Vector Embeddings, and ML Model Weights

When personal data feeds machine learning and AI pipelines, the legal obligations of DPDP propagate through all derived data layers. Under Section 2(t), any derived data from which a natural person can be directly or indirectly identified remains personal data. If a principal withdraws consent under Section 6(4), the enterprise must evaluate the impact across: (1) Raw Feature Stores -> (2) Aggregated Metrics -> (3) Vector Embeddings -> (4) Model Training Checkpoints -> (5) Inference Logs.

Statutory Source: Ch. 21, Topic 10
Archetype: ai reach
I-25 DiagramArchetype: AI REACH

Derived-Data & AI Reach Map

Consent Revocation Propagation across AI & ML Pipeline Stages
Mandatory Purge
Stage 1: Raw Ingestion
Personal Data

Immediate deletion upon s.6(4) revocation.

Automated Prune
Stage 2: Feature Store
Engineered Features

Prune user row from offline and online feature tables.

Vector Eviction
Stage 3: Vector Embeddings
Derived Vectors

Delete user vector ID from Pinecone / Milvus / pgvector.

Retrain Cycle
Stage 4: Model Weights
Trained Model Checkpoint

Machine unlearning or scheduled model retraining if personal data memorized.

Log Masking
Stage 5: Inference Logs
Prompt / Response Logs

Scrub user identity and prompt history from telemetry archives.

💡 Core Architectural Insights & Takeaways

Key technical and regulatory takeaways established by this architecture diagram.

Derived representations that allow re-identification remain personal data under Section 2(t).
Vector embeddings linked to principal IDs must be purged upon consent withdrawal.
Machine unlearning techniques or scheduled retrain cycles are required for LLM fine-tuning datasets.

📜 Statutory Grounding & Legal Perimeter

Primary Act and subordinate Rule provisions establishing the enforceable legal mandate for this diagram.

Enforceable Provisions:
Section 2(t)Section 6(4)Section 8(7)Section 12(3)

🏗️ Target Architecture & Impacted Systems

Enterprise nodes and store topologies impacted by this architectural pattern.

View Complete Topology (SYS-001..014) →

⚙️ Associated Operational Controls

Control Master Matrix obligations enforcing the standards illustrated in this infographic.

Open Control Matrix →

🛠️ Implementation Guidance & Traps

Recommended technical sequencing and operational failure modes to avoid.

Recommended Implementation Sequence

  1. Maintain a Model Data Provenance Registry linking training datasets to consent version IDs.
  2. Build automated vector deletion pipelines on RAG knowledge bases.
  3. Implement differential privacy or strict anonymization prior to feature engineering.

Enterprise Traps & Failure Modes

  • ⚠️ Assuming mathematical embeddings or derived scores are automatically non-personal data.
  • ⚠️ Training foundational models on unconsented user data without retention controls.

📁 Verifiable Evidence Artifacts Vault

Required evidentiary workpapers for regulatory inspections and SDF audits.

📄 AI / ML Data Lineage Manifest
📄 Vector Store Erasure Trigger Verification Log
📄 Differential Privacy Anonymization Certificate