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.

Categories & Use Cases

Media

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.

Microsoft Purview Data Governance

Microsoft 365 Copilot

Untraceable

What is Untraceable?

Categories & Use Cases

Media

Untraceable Alternatives