A customer asks us to delete their data. How do you handle copies in retrieval, logs, backups and trained models?
Make your call first. Use the published answer and rubric to examine your reasoning when you are ready.
Case briefRead the inputs, then make your call
A customer requests deletion from an AI document assistant. This is a fictional case; determine whether the request concerns a tenant, an individual or an optional training use.
| Active inventory | 2,400 derived objects; workers report 2,280 removed and 120 pending. |
|---|---|
| Residual access | Ten pending index entries remain retrievable. |
| Other copies | Source replicas, chunks, embeddings, response caches, diagnostic logs, exports, backups and external processors may contain copies. |
| Model question | Some source data may have entered training datasets used by deployed model versions. Lineage and model retention are not yet established. |
| Customer expectation | The customer wants a completion statement. Legal retention exceptions and the applicable deadline still require confirmation. |
Your task. Choose the immediate action, design a verifiable deletion workflow and write the scope of the status you can truthfully report.
0 words