Data Retention Policies for Chat Assistants
Regulators and customers expect clear retention rules. This guide helps you define per-artifact policies and implement deletion workflows that stand up to audits.
Classify your data
| Artifact | Example | Default retention |
|---|---|---|
| Chat transcripts | Conversations, citations, feedback | 90 days |
| Analytics events | Impressions, opens, fallback logs | 90 days |
| Audit configs | Prompt versions, crawl manifests, threshold changes | 365 days |
| Backups/snapshots | Database or storage snapshots | Per backup policy |
Tenant overrides
- Allow enterprise tenants to extend retention up to 730 days for analytics/audit needs.
- Store retention policies per tenant and per artifact type.
- Reflect retention in your pricing tiers; longer retention impacts storage costs.
Deletion workflow
- Request: Tenant triggers deletion via UI/API; capture reason and scope (single tenant or per user).
- Queue: Add job to compliance queue; include retention policy and deadlines.
- Delete: Remove data from primary databases, search indexes, and backups (or mark for purge).
- Evidence: Store hash of deleted content, actor, timestamp, and scope to prove deletion occurred.
- Notify: Send confirmation to tenant admins and log the event.
Automation tips
- Use time-to-live (TTL) indexes where databases support them.
- Run daily jobs that prune expired documents and emit summary metrics (records deleted, bytes freed).
- Integrate with backups: mark entries for purge and ensure they are removed within backup retention windows.
CrawlBot policy
CrawlBot enforces 90/90/365 defaults, supports tenant overrides up to 730 days, and logs every deletion. Use this framework to keep assistant data compliant and audit-ready.***