CarveLines
Privacy

A privacy layer for sensitive work.

CarveLines works from its own research and the material you choose to share. For sensitive workflows, a local sanitization layer is designed to remove confidential values before external AI processing and restore them locally afterwards.

Light refracted through a glass of water onto a table
Today

What we work with today

  • Our own research: web extraction and data tools
  • Your briefs, notes and reference documents
  • Formatting instructions and output standards
  • Analytical frameworks and house style
The local sanitization layer

Sensitive values stay on the local side of the line.

The layer sits below the products. It is a privacy and context layer, designed to let future workflows use strong external models while reducing the need to expose confidential information to them.

  1. On your computer

    Local intake

    The document is opened locally and its structure, formatting and tables are extracted.

  2. On your computer

    Sanitize

    Sensitive fields are detected with local rules or models and replaced by stable tokens.

  3. External model

    Process

    Only the sanitized content is sent, when external processing is required.

  4. On your computer

    Restore

    The output returns and the original values are reinserted locally.

  5. On your computer

    Assemble

    The final deck or document is rebuilt with its structure and formatting intact.

Stable tokens keep the reasoning intact

Each sensitive value is replaced by the same neutral token everywhere it appears in the file, so structure and reasoning survive the round trip.

€47.3m revenue[REVENUE_01]

Project Alpine[PROJECT_01]

What the layer is designed to detect

  • Company names
  • People
  • Amounts
  • Internal KPIs
  • Deal names
  • Customer and supplier names
  • Other configured confidential fields
Onboarding service

We capture your context before the first job.

  1. 01A working session to collect your templates, brand chart, reference deliverables and review rules
  2. 02Context configured per use case: PE screening, IC deck, board pack, investor day
  3. 03First deliverable reviewed together; the context is refined after every job

Your context is used only by you. Sensitive values are sanitized locally before any external AI processing, and each client's context is stored separately.

Availability

The layer is enabled workflow by workflow, as each one is ready. Tell us about yours and we will say what is possible today.

Have a sensitive workflow in mind?

Tell us about it. We will say what is possible today.

Request access