Secured AI

Secured AI is a privacy layer that detects sensitive data in a prompt and masks it before it reaches an AI model, then restores the original values locally so the output remains usable.

Key features and functions include:

Real-Time Detection and Masking
Sensitive data entered into an AI prompt is identified and replaced with context-preserving placeholders before the prompt reaches the model, so the underlying request structure is retained without exposing the original values.

Legal-Specific Data Detection
Detection is tuned for legal documents, case citations, and privileged-content markers, covering client information (names, contact details, matter numbers, billing information), case details (case names, court information, opposing parties, witness names), confidential documents (contract terms, settlement details, intellectual property information, trade secrets), and privileged communications (legal advice, case strategy, mental impressions, work product).

Reveal Technology
Masked values in the AI's response are restored locally on the user's device using a Master Key that the user controls, so the output displays real names, terms, and figures rather than placeholder text.

Matter-Level and Role-Based Controls
Protection policies can be configured at the matter or client level, and access to reveal operations is role-based across attorneys, paralegals, and other staff.

AI Usage Records
The product logs what data was sent to an AI model, what was masked, and what was returned, which firms can use for internal ethics review or when billing AI-assisted work.

Model Support
Secured AI currently integrates with OpenAI and DeepSeek models.

Zero-Knowledge Architecture
Data protection is built on a zero-knowledge vault architecture with local encryption; the company states it does not store or have access to a user's Master Key.