Security
SupportAI is built with production-minded controls for authentication, data separation, anti-abuse checks, and safe operational visibility.
Security Control Lab
Risk score
62
Blocked attempts
112
Escalations
19
Server-side validation and typed mutation schemas
Every incoming action is validated before processing and rejected when payload shape is unsafe.
Status: hardened
Role-based authorization and workspace scoping
Routes and mutations enforce role checks and workspace boundaries to prevent horizontal data access.
Prompt guardrails and low-confidence escalation routing
Unsafe or uncertain assistant responses are downgraded and routed into human escalation workflows.
Status: stable
Rate limiting architecture for auth and chat submissions
Abuse thresholds are applied at request edge and sensitive endpoints to contain automated attack patterns.
Audit logging boundaries for sensitive configuration changes
Configuration updates are logged with actor context to improve forensic review and compliance visibility.
Status: monitoring
Safe rendering with sanitized text input pathways
User content is normalized and sanitized before render so script injection vectors are blocked by default.
Future roadmap includes anomaly detection, provider-side moderation, retention policy controls, and encrypted sensitive fields.