Cloaking replaces every confidential value with an informative, meaning-preserving cloak on your own computer before any AI model sees the text. VaultScribe is the field-expert writer built on that engine; it drafts, rewrites and reviews Pharma, HEOR, RWE, regulatory and biostatistics documents at near-frontier quality without the model ever receiving a real value.
Cloaking: the engine every tool runs on
Most "compliant AI" protects you with a contract: a BAA, a no-training promise, a data-processing addendum. The confidential value still reaches the model; you are simply trusting a policy about what happens to it afterward. Cloaking protects you with architecture: the value never reaches the model in the first place.
Before any text leaves your computer, Cloaking finds each sensitive value and replaces it with an informative cloak, not a black box like [REDACTED], but a label the model can reason over. The model reads a coherent, fluent document; it simply never learns the real names, numbers or terms behind the cloaks. When the response comes back, the truth is re-inserted on your screen only.
Advanced (semantic) Cloaking retains 94–98% of the model's natural writing quality; Rapid (sequential) Cloaking retains 91–93%, versus roughly 32–37% for censoring and substitution-style redaction.
Source: Tremblay & Harricharan, 2026, Comparing De-identification Methods on AI Writing Quality in HEOR and Market Access (doi:10.5281/zenodo.21343321), rebalanced out of the study’s 96/100 baseline. The ~37% and ~32% de-identification figures are provisional pending publication of the per-condition data.
Two tiers, one guarantee
- Rapid Cloaking: fast sequential cloaking for quick drafting and conversational work; 91–93% writing-quality retention against the uncensored baseline.
- Advanced Cloaking: semantic cloaking that preserves the meaning and relationships behind each value; 94–98% retention for high-stakes documents.
- Either way, the AI only ever sees cloaks. Real values never leave your environment, and every submission is auditable: the audit trail records the cloaks and run IDs, never a real value.
VaultScribe: a field expert, not a generic chatbot
A frontier model on its own is a fluent generalist. VaultScribe turns it into a specialist for regulated work. It combines Cloaking with Hybrid RAG and a Grounding Codex (a field-specialized, human-anchored knowledge base), so drafts are grounded in real domain references rather than the model's unaided priors.
- Draft, produce first-pass HEOR dossiers, regulatory sections, value narratives and biostatistics text from your own cloaked source material.
- Rewrite: tighten, restructure or re-tone existing content while keeping every confidential value cloaked end to end.
- Review, get expert-level critique and consistency checks across a document, again without exposing a single real value.
- DeCloak, receive the finished document back in SharePoint in the identical format as the original: figures and tables preserved, cover and back pages intact, and every cloaked value restored to its real value, so you see the truth.
The result is the writing quality of a frontier model with the confidentiality posture of a tool that never received your secrets, which is exactly what NDA-bound, regulated work requires.
The cloak underneath
Every Untraceable tool runs on the same patent-pending cloaking engine. Confidential values (PII, PHI, NDA-bound and trade-secret data) are replaced with informative, meaning-preserving cloaks on your own computer, before any AI model sees the text. The model reasons over cloaks and never receives the real value; the truth is re-inserted only on your screen, after the model responds. That is what makes every product on this page safe to point at data you can't share. How cloaking works →
Frequently asked questions
What is Cloaking?
Cloaking is Untraceable's patent-pending method for replacing each confidential value (a name, a price, an NDA-bound term) with an informative, meaning-preserving cloak before any AI model sees the text. Two tiers: Rapid (sequential) Cloaking retains 91–93% of the model's natural writing quality, and Advanced (semantic) Cloaking retains 94–98% (Tremblay & Harricharan, 2026, doi:10.5281/zenodo.21343321). Unlike redaction, the model keeps the context it needs to write well; unlike encryption, the cloak is something the model can reason over.
What is VaultScribe?
VaultScribe is a field-expert AI writer built on top of Cloaking. It drafts, rewrites and reviews regulated documents (HEOR dossiers, regulatory submissions, biostatistics narratives) using Hybrid RAG and a human-anchored Grounding Codex for domain accuracy. Because it runs on cloaked data, you get frontier-model writing quality without ever sending the confidential value to the model.
Does the AI ever see my real data with VaultScribe?
Never. Cloaking runs client-side, in your browser, before anything is transmitted. The AI receives only cloaks; your files and real values stay in your own environment. VaultScribe re-inserts the truth on your screen after the model responds; we never see your real values or your keys, only the cloaked text on its way to the model.
How is VaultScribe different from a de-identification API plus ChatGPT?
De-identification strips context, which guts writing quality and invites hallucination, and it protects persons, not the information an NDA covers. VaultScribe's cloaks preserve meaning, so writing quality stays near-frontier, and it is purpose-built for the document types regulated experts actually produce, with an audit trail of every cloak and every AI submission.
Which fields is VaultScribe fluent in?
Health economics and outcomes research (HEOR), market access, regulatory affairs, medical writing and biostatistics today, with legal, financial and professional services, and management consulting, expanding. The Grounding Codex anchors drafts to real, field-specific references rather than the model's unaided priors.
Work with AI on data you can't share with it.
Untraceable cloaks every confidential value on your computer before any AI model sees the text; the model never receives the secret at all.