Security First

Your AI avatars handle sensitive company communications. We built Surrogate with enterprise-grade security from day one.

Encryption

AES-256-GCM encryption for all stored credentials
Separate encryption keys per organization
OAuth tokens encrypted at rest, decrypted only at runtime
TLS 1.3 for all data in transit

AI Sandboxing

Closed tool whitelist — no arbitrary code execution
No bash, file system, or network access for AI agents
Input sanitization strips prompt injection attempts
Output filtering redacts leaked credentials and PII
Hard timeout (30s) on all AI operations
Per-avatar rate limiting on LLM calls

Data Isolation

Multi-tenant architecture with organization-level isolation
Row-level security in PostgreSQL
Each avatar scoped to its own integrations only
API keys never passed into LLM context or prompts

Audit & Compliance

Complete audit logging for all avatar actions
Conversation data retention configurable per org
GDPR-compliant data export and deletion
SOC2 readiness from day one

Prompt Injection Defense

Multi-layer input sanitization
System prompts include anti-injection instructions
Real-time output monitoring for data leakage
Automated detection of extraction attempts

Questions about security?

Contact us for a detailed security whitepaper or to discuss enterprise requirements.