Overview
Detect toxicity/PII/jailbreaks; block, strip, or redact; and keep auditable logs.
Problem
Unfiltered AI and open-text chats risk abuse, leakage, and regulatory violations.
Solution
Chat Core enforces input/output rules across detectors and LLMs: redact PII, downgrade confidence, or block content; keep immutable audit trails.
How it works
Enable moderation categories; configure redaction masks for PAN/SSN/email/phone; require citations for KB answers; export block events to your SIEM.
Who is this for
Security
Compliance
CX Platform Owner
Expected outcomes
- Lower policy violations and data leaks
- Confidence bands on AI replies with traceability
Key metrics
Blocked/stripped messages
Baseline
0 per 1k
Target
15 per 1k
PII in transcripts
Baseline
450 ppm
Target
20 ppm
Gallery
Downloads & templates
Case studies
Healthcare marketplace adopts strict masking
PII leakage in transcripts dropped by 95%.
Healthcare Enterprise NA
Security impact
- Full transcripts (masked), decision traces · PII: minimized via redaction
Compliance
- GDPR (data minimization & rights)
- HIPAA add-on (BAA)
- SOC2 (security, confidentiality)
Availability & next steps
Business
Enterprise