Legal
Security
Transparent security practices. Current status.
Privacy-first architecture
Seer's behavioral analytics system is designed to never collect personally identifiable information. This is not a policy choice. It is a technical architecture decision. Contact data is only captured when a visitor voluntarily provides it through the AI chat interface, and is stored in an isolated data path separate from analytics.
- Anonymous behavioral patterns
- Session IDs (random, non-reversible)
- Viewport dimensions
- URL paths (no query params)
- Timestamp metadata
- Name, email, phone
- Only when visitor provides it via chat
- Stored securely, encrypted at rest
- Used only to deliver contracted services
- Never sold or shared
- PII from analytics (fully anonymized)
- IP addresses
- Payment information
- Browser fingerprints
- Cross-site tracking
Infrastructure security
Hosting and storage
- AWS: Encrypted infrastructure (us-east-1)
- Encryption: AES-256 at rest
- Database: Supabase PostgreSQL with RLS
- Backups: Daily automated, 30-day retention
Transmission security
- TLS 1.3: All endpoints
- CDN: Cloudflare with DDoS protection + WAF
- WebSockets: Secure WSS protocol
- IP Anonymization: Cloudflare strips IPs
Access control
Authentication
- User Auth: Clerk-based with MFA support
- API Keys: Tenant-specific, rotatable
- Webhooks: HMAC-SHA256 signatures
Internal access
- Role-based access control (RBAC)
- Audit logging for all data access
- Production access: senior engineers only
- No direct database access
AI security
Seer's AI assistant is hardened against prompt injection, social engineering, and data exfiltration attacks with defense-in-depth architecture.
Prompt security
- Server-Side Only: System prompts never touch the browser
- Injection Defense: Multi-pattern detection for manipulation attempts
- Input Sanitization: All user messages scanned before AI processing
- Security Logging: Flagged attempts logged for review
Honeypot monitoring
- Canary Phrases: Unique strings embedded in system instructions
- Output Scanning: Every AI response checked for leaked phrases
- Instant Alerts: Security team notified on any detection
- Severity Tiers: Critical, high, medium classifications
Rate limiting
- Per-Session: 30 messages maximum per session
- Per-Minute: 7 messages per minute burst limit
- Per-IP: 10 requests per minute per IP address
- Auto-Cleanup: Rate limit data purged every 5 minutes
Tenant isolation
- Knowledge Base: Strict tenant_id filtering on all queries
- No Cross-Tenant Bleed: Each tenant's AI is isolated
- Session Binding: Conversations locked to originating tenant
Data retention
All behavioral session data is automatically and permanently deleted after 30 days. Contact data provided through the chat interface is retained for the duration of the client relationship and deleted within 30 days of contract termination.
- Session telemetry
- Emotional detections
- Intervention logs
- Aggregated stats only
- Anonymized ML data
- Contact data (client relationship)
- Billing info (legal req)
Security roadmap
Incident response
- 01Detection: Automated monitoring detects anomalies
- 02Containment: Immediate isolation of affected systems
- 03Investigation: Root cause analysis within 24 hours
- 04Notification: Affected customers notified within 72 hours
- 05Remediation: Fixes deployed and verified
- 06Post-Mortem: Public incident report published
Transparency commitment
- ✓No marketing fluff: We tell you exactly what we do
- ✓No false claims: Certifications listed when obtained
- ✓Public incident reports: All security incidents documented
- ✓Open architecture: Technical docs available for review
Security Contact
For security inquiries, vulnerability reports, or compliance documentation:
security@sentientiq.appResponsible disclosure is appreciated and will be acknowledged publicly (with your permission).