TrustRadius: an HG Insights company

Untraceable

What is Untraceable?

Untraceable (developed by Intraable AI Inc.) is an artificial intelligence (AI) workspace and Data Masking Software platform engineered for confidential and contractually restricted workflows. Designed for consultants, biostatisticians, and regulated-industry analysts operating across Health Economics and Outcomes Research (HEOR), pharmaceuticals, legal, and regulatory affairs, the software applies client-side semantic cloaking to text prior to model transmission.

Architectural Privacy & Semantic Cloaking

Rather than relying solely on contractual non-training agreements or network perimeter security, Untraceable replaces confidential values—including Personally Identifiable Information (PII), Protected Health Information (PHI), Non-Disclosure Agreement (NDA) client records, pricing models, and unpublished research results—with semantic cloaks directly on the local endpoint before data leaves the user's environment.

These semantic cloaks preserve grammatical type, units, and context within the sentence, allowing large language models to reason about variables without receiving raw underlying data. Original values are restored locally on the user's display. According to vendor-cited benchmarks, semantic cloaking maintains high output quality relative to uncloaked text baselines while expanding coverage beyond fixed identifier lists such as Health Insurance Portability and Accountability Act (HIPAA) categories.

System Auditability & Suite Architecture

To support compliance reviews and client audits, every AI interaction generates an audit trail that logs cloak labels and execution identifiers without storing underlying raw values. VaultAudit provides deliverable oversight and issues compliance verification against the Untraceable Protocol specification.

The platform suite includes specialized interaction modules:

  • VaultScribe: A document drafting add-in integrated into Microsoft Word and PowerPoint that applies paragraph-level and deck-wide cloaking with tracked revisions.
  • VaultChat: A conversational AI interface for rapid document querying.
  • VaultDelphi: A multi-model consensus tool for querying multiple frontier models simultaneously.
  • Blueprint & VaultCanvas: Structured templates and automated tools for client reporting.
  • VaultStat: A statistical analysis tool engineered for cloaked dataset processing.

Deployment & Availability

Untraceable operates as a model-agnostic platform supporting web applications and Microsoft 365 integrations (Word, Excel, PowerPoint). The software is available via the Microsoft Commercial Marketplace with billing integrated into existing Azure accounts. Data residency is maintained in the United States and Canada, with user interfaces supported in English and French.

Categories & Use Cases

Screenshots

Screenshot of Screening every detected value before anything is sent. Each candidate is listed with its surrounding sentence so the user can confirm it in context, accept the proposed sequential cloak, or clear it as non-confidential. The counter tracks progress against the full set, here 81 of 81 value groups reviewed across sequential, semantic and name categories. Nothing is auto-cleared; the decision stays with the operator.
Screenshot of A project folder holding original and cloaked documents side by side. Word, Excel and PowerPoint files are cloaked in place and versioned, so the cloaked copy that was sent to the AI remains available next to the original for audit. The confidentiality suite across the top covers cloaking, the Untraceable Protocol audit and certification view, and VaultScribe for AI drafting.
Screenshot of What the user sees versus what the AI receives. The same cost-effectiveness table, shown twice: on the left the real values and the Ontario Case Costing Initiative source, on the right the version transmitted to the AI model, where every figure has been replaced with a semantic cloak such as SLC_0234 and the location with NAME_LOCATION_1. Structure, units, footnotes and the relationships between cells are preserved, which is why the model can still reason about the table.
Screenshot of A VaultDelphi report. Several frontier models answer the same question blind, self-score their own evidence, and the panel runs further rounds until convergence or a documented failure to converge. Here the panel did not reach consensus on an incidence figure, and the report says so, naming the methodological split and the sources behind each position rather than returning a single confident number.

1 / 4

Screenshot of Screening every detected value before anything is sent. Each candidate is listed with its surrounding sentence so the user can confirm it in context, accept the proposed sequential cloak, or clear it as non-confidential. The counter tracks progress against the full set, here 81 of 81 value groups reviewed across sequential, semantic and name categories. Nothing is auto-cleared; the decision stays with the operator.

Technical Details

Technical Details
Deployment TypesSaaS
Mobile ApplicationNo
Supported CountriesCanada, United states
Supported LanguagesEnglish, French

FAQs

How much does Untraceable cost?
Untraceable starts at $25.
What are Untraceable's top competitors?
Microsoft Purview Data Governance and Microsoft 365 Copilot are common alternatives for Untraceable.
Who uses Untraceable?
The most common users of Untraceable are from Small-Business.