Security & Trust

• Segregated Environments: Production and staging environments are completely isolated
• Least-Privilege Access: All system access follows principle of minimum required permissions
• Role-Based Permissions: Human agents have granular, auditable access controls
• Immutable Audit Logs: All critical actions are logged and cannot be modified

• Encryption in Transit: All data transmission uses TLS 1.2+ protocols
• Encryption at Rest: Stored data is encrypted using industry-standard algorithms
• Hashed Identifiers: Sensitive identifiers are tokenized and hashed
• Secure Key Management: Cryptographic keys are rotated and stored securely

• Persona Isolation: AI personas are strictly enforced and isolated
• Escalation Thresholds: Confidence-based triggers for human review
• Human Review Gating: Critical decisions require human approval
• AI Whisper Mode: AI provides suggestions without taking control

• Rate Limiting: Per embed token and session rate limits prevent abuse
• Email Cooldowns: Escalation requests have built-in cooldown periods
• Prompt Injection Detection: Advanced filtering prevents malicious prompts
• Anomaly Detection: Automated systems identify and flag suspicious behavior

• 24/7 System Monitoring: Continuous monitoring of all critical systems
• Security Alerts: Real-time alerting for security events
• Incident Response Procedures: Documented and tested response protocols
• Post-Incident Review: Comprehensive analysis and remediation after incidents

SiskoAI relies on carefully vetted third-party providers for AI processing, payments, and infrastructure. All vendors undergo rigorous security and compliance reviews.

• Security posture assessments for all critical vendors
• Compliance verification and documentation
• Regular vendor security reviews and audits
• Contractual security requirements and SLAs