How It Works
How SecuredAI Protects Your Data
Three capabilities working together: detect sensitive data, obfuscate it with context-preserving entities, and restore it using your zero-knowledge vault. That last part is what makes SecuredAI different.
The Three-Part Protection Pipeline
Every AI interaction passes through three stages: Detection catches the sensitive data, Obfuscation protects it, and the Zero-Knowledge Vault restores it for you.
Stage 1
Detection
Identifies multiple PII types before sensitive data is processed. Names, emails, phone numbers, and more are caught instantly.
- Real-time identification of sensitive data
- Catches data before it leaves your device
- Multiple data types recognized automatically
Stage 2
Obfuscation
Sensitive data becomes system-generated entities (e.g., [PERSON_A]). Entities retain utility but protect the original values.
- Names become [PERSON_A], emails become [EMAIL_1]
- AI sees the structure, not the sensitive values
- System-generated entities preserve meaning
- Entities not automatically detected can be manually added, giving you full control over what gets obfuscated
Stage 3
Zero-Knowledge Vault
Original data is restored locally on your device. Every interaction is tracked and recorded.
- Restoration happens on your device
- Each user sets up their own unique Master Key during onboarding
- SecuredAI never stores or has access to your Master Key
- SecuredAI never sees your sensitive data
Core Capabilities
What Gets Detected
Names, emails, phone numbers, organization names, account numbers, and identifiers. Built to handle private information across use cases.
How Obfuscation Works
System-generated entities are context-aware to maintain 'data sense.' The result is seamless AI accuracy even when the source data is protected.
What Zero-Knowledge Vault Does
Ensures sensitive data is visible only to you when you need it. No manual lookups. No leaks of sensitive information. Data stays on your device.
The Missing Piece
Why the Zero-Knowledge Vault Changes Everything
The Problem with Obfuscation Alone
Other AI privacy tools stop at obfuscation. They detect sensitive data, replace it with placeholders, and send protected prompts to the LLM. But here's what happens next:
- ×AI responses come back full of [REDACTED] placeholders
- ×Users can't act on answers they can't understand
- ×Manual reconstruction wastes time and introduces errors
- ×Teams abandon the tool entirely—or bypass protection
How the Zero-Knowledge Vault Solves This
SecuredAI stores an encrypted mapping on your device. When the LLM response returns, the Zero-Knowledge Vault automatically restores the original values— locally. SecuredAI never sees your sensitive data.
Without Zero-Knowledge Vault
"[PERSON_A] should schedule a follow-up with [PERSON_B] at [ORG_1] within 2 weeks of [DATE_1]."
Unusable. Who is the person? Which organization? What date?
With Zero-Knowledge Vault (on your device)
"John Smith should schedule a follow-up with Dr. Sarah Chen at Boston Medical Center within 2 weeks of March 15, 2024."
Actionable. User knows exactly what to do.
Works With OpenAI and DeepSeek
Deploy protection with the AI models we currently support. More models coming soon.
Security by Design
SecuredAI was built from day one with security-first principles.
Zero-knowledge vault
Your data is never stored on our servers
Local encryption
Your Master Key encrypts everything locally
Master Key is never stored server-side
Only obfuscated content ever transmitted