Mask PII before the LLM. Zero retention is a toggle away.
Indexed content and questions from the widget, API, and Ask can be scanned for PII (emails, names, phone numbers, card numbers, IBANs) and masked before they reach any model. Workspaces can go further and enable zero retention — questions and answers are streamed and dropped, never persisted. PII masking is switched on per knowledge base, zero retention per workspace.
What PII masking + zero retention actually do
Two controls, deployed correctly — and audit-friendly
PII Masking
- Detects emails, phones, credit cards, IBANs, IP addresses, titled names
- Applied to widget, API, and Ask questions before any LLM call — including reranking
- Applied before storage — the DB holds the masked form
- Custom regex patterns and allow-list terms per knowledge base
- Cited answers still reference the underlying doc, not the raw user question
Zero Retention
- Per-workspace toggle: questions + answers stream but never persist
- No conversation row is created; no feedback can be attached
- Still generates cited, grounded answers in real time
- Compatible with SSO, restricted documents, and per-KB permissions
- Still counts toward the monthly quota, as a daily count with no text
Compliance Posture
- Workspace-scoped data isolation across content, embeddings, and chats
- Encryption in transit and at rest
- Your data is never used to train models — none of it, ever
- Data-processing agreement (DPA) part of the Terms on every plan
- Data stored in the United States, with SCCs for EEA, UK and Swiss data
The two questions every security review asks
'Does user PII reach a third-party model?' and 'Where does the conversation get stored?' are the two questions that stall AI procurement. PII masking answers the first before it can be asked. Zero retention answers the second — for workspaces that choose it, the answer is 'nowhere; it streams and it's gone'. Both are architectural controls, not aspirational policy.
- Masking runs before the LLM calls for widget, API, and Ask questions: chat and rerank
- Zero-retention workspaces skip persistence entirely — no message row exists
- Audit logs cover admin actions; traces record each answer
- SSO (OIDC + SAML) and restricted docs stack on top
- Enterprise adds a countersigned DPA and dedicated support
Asked in the widget
Hi, I’m Dr. Priya Raman. Where is order ORD-48213? Reply to [email protected].
What the model sees
Hi, I’m [NAME]. Where is order [REDACTED]? Reply to [EMAIL].
PII Masking
Enable PII masking
Custom regex patterns
\bORD-\d+\b
Data Retention
Warning: with zero data retention, conversations and traces are not stored at all.
Private content stays private, and secrets stay sealed
Private stays private
Internal knowledge bases have no public surface at all. To keep past tickets private, add them to an internal knowledge base.
Credentials are encrypted
Connection secrets are sealed with AES-256-GCM. If one can't be opened, the sync stops instead of running without it.
Every answer from your docs shows its sources
Numbered links point to the page and section each part of the answer came from, so anyone can check it in one click.
Got questions?
Frequently Asked Questions
Common questions about PII & Zero Retention
Emails, phone numbers, credit-card numbers, IBANs, IP addresses, and personal names that follow a title such as Dr. Any plan can add custom regex patterns per knowledge base for domain-specific identifiers.
Once you switch it on for a knowledge base: before any LLM call and before storage. Document bodies are masked before they are stored or embedded. Questions from the widget, the public chat API, and Ask are masked before the chat model or the reranker sees them. Questions from bots, the form deflector, helpdesk automation, and MCP are not masked.
Questions and answers stream to the caller in real time and are then dropped — no conversation row is created, no message is stored, no feedback can be attached. Retrieval still happens against your indexed content; the model still generates a cited answer. Only the transcript is not persisted.
Yes. Zero-retention questions still count toward the monthly AI-question quota: Kelu keeps a count of questions per day, with no text. Enterprise contracts price this differently; see pricing for details.
No. Indexed content, embeddings, chats, and feedback stay isolated to your workspace and are never shared across customers or used to train any model — ours or a third party's. This is contractual on paid plans and technical on every plan.
The providers Kelu uses run under zero-retention API terms (no training on API inputs) as documented in their enterprise agreements. You can bring your own model credentials (BYOK) so the traffic runs on your own account, under your own terms.
Bring this to your security team
Read the security overview and the DPA, or talk to us about a countersigned copy on Enterprise.
Free to start · No credit card required