Trust Center: How UltraProbe Handles Your Data
Data handling
Prompt scans are analyzed in-memory and immediately discarded — never stored, logged, or sent to third parties. Website scans fetch your public HTML and discard the raw content after analysis; reports keep only the scan results (grade, score, categories), never the original content. Email is collected only when you voluntarily submit it to unlock a full report, with one-click unsubscribe.
Security practices
TLS 1.3 in transit (HTTPS only). 90% of analysis is pure regex — SEO/AEO scans make no LLM API calls. Per-IP rate limiting, input sanitization against SQL injection and XSS, no tracking cookies, no fingerprinting.
Infrastructure
Hosted on Vercel (Edge Network), database on Firebase Firestore (asia-east1), email via Resend. Gemini Flash is used only for optional deep analysis. The scanner engine is open source under MIT: github.com/ppcvote/prompt-defense-audit.
Standards and methodology
Full OWASP LLM Top 10 coverage. The AI security scan covers 25 defense vectors (12 LLM-era + 13 agent/ASI-era). SEO: 30+ checks in 8 categories; AEO: 32+ checks in 8 categories.