Coming soon from LegalOps

SafeRoom

A local, customer-aware confidentiality layer for preparing legal documents for external AI analysis.

Protect the client. Preserve the contract.

See how it works Back to LegalOps
Why SafeRoom

The problem begins before the AI prompt.

A legal document can reveal a client through far more than the visible party name. Names, email domains, signature blocks, comments, tracked changes, metadata, thumbnails, project names and contextual clues can all carry identity into an external AI workflow.

01

Protect the customer first

SafeRoom starts with the protected client, then classifies counterparties, people, affiliates and other entities relative to that customer.

02

Preserve legal meaning

Contract structure, clause wording, commercial terms, comments and tracked changes remain useful for analysis wherever possible.

03

Local by design

Document processing is designed to happen on the user's computer, with the original never overwritten and uncertainty shown for review.

The core experience

Five simple steps, almost no technical decisions.

The normal workflow is intentionally minimal. SafeRoom infers what it can, asks the user to confirm the customer, and only surfaces uncertain items that need attention.

Open or drop

Start with a DOCX or supported PDF.

Confirm

Answer the core question, who are we protecting?

Analyse

Run local detection across text, identifiers and hidden document artefacts.

Review

Check only uncertain items and contextual clues that remain.

Use

Copy Safe Text or save a sanitised working copy and audit.

Customer-aware renaming

Semantic replacements keep the relationships intact.

Instead of replacing everything with generic labels, SafeRoom uses role-aware names that preserve the difference between the customer, counterparty, affiliates and other actors.

Protected client organisationCUSTOMER COMPANY
Protected client employeeCUSTOMER PERSON 1
Principal other contracting partyCOUNTERPARTY COMPANY
Counterparty employeeCOUNTERPARTY PERSON 1
Other organisationTHIRD PARTY COMPANY 1
Sensitive project or matter[MATTER IDENTIFIER]
Designed for legal documents

SafeRoom looks beyond visible body text.

The Word workflow is designed around the places confidential identity actually survives in negotiated documents.

Body text, tables, headers and footersNames and identifiers are detected across the document, including text split across formatting runs.
Footnotes and endnotesNegotiated and explanatory note text is included in the sanitisation process.
CommentsComment text can be preserved while reviewer identity and identifiers inside comments are pseudonymised.
Tracked changesInsertions and deletions can remain legally useful while revision text and author traces are sanitised.
Metadata and document propertiesAuthor, company, manager, reviewer, custom properties and non-essential revision traces are scrubbed.
Hidden and visual risksThumbnails are removed, generic output names are used, and unsupported images or embedded objects are flagged for review.
Beyond PII

Could someone still work out who the customer or matter is?

SafeRoom treats contextual confidentiality as a separate risk. Project codenames, matter numbers, internal URLs, customer-specific systems, unusual locations and identifying transaction descriptions can all survive ordinary anonymisation.

Customer protection check

Known customer identifiersCLEAR
Personal identifiersCLEAR
MetadataCLEAR
File and package namesCLEAR
Possible contextual cluesREVIEW
First release

One problem, solved extremely well.

The initial release is focused on preparing Word agreements for external AI review in a few clicks, completely locally. PDF text extraction follows, with OCR and full PDF redaction treated as later controlled phases.

Windows desktop applicationDrag and drop DOCX support with local customer profiles and customer selection.
Role-aware de-identificationCustomer-centric replacements, legal-context detection and an uncertainty review queue.
Useful outputsCopy Safe Text, a sanitised DOCX option, a sanitisation audit and a second-pass residual scan.
Important: SafeRoom is designed as a risk-reduction and de-identification tool, not as a promise of perfect anonymisation. It should surface uncertainty, make residual review easy and create a defensible process rather than claim absolute safety.
Coming soon

SafeRoom will become part of the wider LegalOps product suite.

The longer-term direction includes Word integration, broader PDF support, policy profiles, batch processing, controlled intake and optional Master Contract Log integration.

Back to LegalOps