How it works

A privacy workflow built around review and verification.

Soveau is being designed to help professionals find sensitive information in their documents, decide what should be protected, create a processed Safe Copy, and verify the result before it is shared. Here is the intended workflow, step by step.

Join the PilotSee the eight steps

Prelaunch: this is the planned workflow. Soveau is in development and isn’t generally available yet.

The workflow

Eight steps, from opening a document to a verified Safe Copy.

Every step exists for a reason. Selecting the text to remove is only part of it; the steps before and after matter just as much.

  1. 01Find and reviewImportChoose local documents. Originals stay untouched.
  2. 02Find and reviewDetectLook for sensitive information by category.
  3. 03Find and reviewReviewSee findings in context and judge them.
  4. 04Find and reviewApproveDetected does not mean changed. You decide.
  5. 05Protect and sanitizeRedact or PseudonymizeRemove approved information, or replace it with substitutes.
  6. 06Protect and sanitizeSanitizeAddress hidden information that can survive visible edits.
  7. 07Verify and exportVerifyCheck the Safe Copy independently before it is released.
  8. 08Verify and exportExport Safe CopyA new file, exported separately from the original.

Step by step

What happens at each step.

Find and review

Step 1 of 8

Import

Choose documents from your own computer. Your originals stay untouched.

You select the documents you want to prepare, whether that is one file or a whole set. Soveau creates a privacy job around them and treats the files you chose as read-only sources: nothing is changed in place.

  • One file, a full client folder, or a mixed batch can sit in a single privacy job.
  • Originals are never modified or overwritten.
  • Importing is intended to happen on your device. It is not a cloud upload.

Step 2 of 8

Detect

Soveau looks for sensitive information by category.

Each document is analyzed for the categories of sensitive information set in your Privacy Profile, such as names, addresses, contact details, and identifiers. Soveau is designed to read the text already in a document and, where a page is a scan or an image, to use OCR on your device.

  • Findings are grouped by category and shown where they occur, ready for review.
  • Detection is assistance, not authority. It can miss things or flag things that don’t matter, which is why review comes next.
  • Scanned pages, images, and mixed documents are part of the design.

Detection results are designed for human review. Performance figures will be published only after they’ve been validated.

Step 3 of 8

Review

You see what was found, in context.

Findings appear where they occur in the document, so you can judge them properly. Clear, repetitive findings are intended to be handled together, while uncertain or important ones are meant to be surfaced for closer attention rather than quietly cleared.

  • You can accept or reject what Soveau found, and add items it missed.
  • Uncertainty is shown, not hidden.
  • The decisions stay with a person.

Step 4 of 8

Approve

Detected does not mean changed.

Nothing is transformed simply because it was detected. You decide what should be protected, and only approved findings move on. Grouped approval is intended to reduce repetitive work when the same kind of value appears many times.

  • Approval is an explicit step, not an automatic side effect of detection.
  • Similar findings can be approved together.
  • Approved values are then redacted or pseudonymized.

Protect and sanitize

Step 5 of 8

Redact or Pseudonymize

Two ways to protect approved information.

Soveau is designed to offer two different treatments for approved findings. They do different jobs, and it is worth choosing deliberately.

Redact

Permanently removes the approved sensitive information from the Safe Copy.

Best when the information isn’t needed at all in the shared document.

Pseudonymize

Replaces approved values with controlled substitutes, keeping the document useful where that is appropriate. For example, the same person can be represented consistently throughout.

Useful when readers still need to follow who or what is being referred to.

Pseudonymization isn’t right for every sharing situation. Substitutes can still allow inference in some contexts, so choose it on purpose, not by default.

Step 6 of 8

Sanitize

Covering text on the page isn’t enough.

Visible changes don’t always remove everything. Depending on the file format, information can also survive in places you don’t see. Soveau’s sanitization step is intended to address that hidden material, such as document metadata and other hidden content, before the Safe Copy is finalized.

  • What needs attention depends on the format of the document.
  • Sanitization is a separate step from redaction on purpose.

Verify and export

Step 7 of 8

Verify

A first-class step

The Safe Copy is checked on its own.

After the Safe Copy is written, Soveau is designed to reopen it and check it separately from the step that created it, looking for protected information that may remain. If a check fails or can’t be completed, the result is meant to be flagged rather than treated as safe.

  • Verification is independent of the transformation that produced the file.
  • Reviewing the outcome is part of releasing it.
  • No check is infallible, so verification adds to human review rather than replacing it.

Step 8 of 8

Export Safe Copy

A new file, exported separately.

The result is a new document. Your original stays exactly as it was, and you decide where the Safe Copy goes next.

  • The Safe Copy is a derivative, saved separately from the original.
  • You choose where to save it and who to share it with.

Batch privacy

Built for more than one file at a time.

A single privacy job can cover a whole set of documents. Soveau’s batch approach is designed around three common shapes of work.

  • One client across many files

    Review findings across a whole client folder together, instead of file by file.

  • Many clients with separate files

    Work through documents belonging to different clients, with each client’s files kept separate and each Safe Copy produced on its own.

  • Mixed documents and entities

    Digital PDFs, scans, and photographed pages side by side, with different people, companies, and identifiers among them.

Privacy Profiles follow categories, not one name.

A Privacy Profile describes the kinds of sensitive information to look for, not a single value to search for. Different documents can contain different values while following the same profile. Categories can include:

  • Names
  • Addresses
  • Email addresses
  • Phone numbers
  • Dates of birth
  • Account, client, and reference identifiers
  • Other configured sensitive fields

Try the workflow early, or help shape it.

We’re forming an early pilot group. If you’d rather start by telling us what slows you down, see Shape Soveau. Submissions aren’t open yet, and we can’t promise dates.

Join the PilotShape Soveau