Verifying Training-Data Erasure

Walks a Data Subject Request (DSR) erasure verification across the four stages of an ML pipeline where a subject's data can persist — raw/curated datasets, derived embeddings or representations, fine-tuned model weights, and inference-time caches or generated content — and produces a tamper-evident proof bundle (per-stage erasure evidence with timestamps, tool output, residual-risk acknowledgement). Use when an organization receives a GDPR Article 17 / CCPA / HIPAA right-to-be-forgotten request that touches data already used in training or retrieval, when an auditor asks for proof that a deleted user's data is gone from a deployed model, or when a fine-tuned model is bein…