# Sidian Technologies Corp. · DataGuard

> DataGuard is secure file sharing with the compliance check built in, made for law firms. Before a document or email leaves the firm, it runs the firm's pre-flight checklist over every page, removes what the recipient is not entitled to see, cites the rule behind each removal, and records who approved it. Files never move out of Clio, SharePoint, OneDrive, Google Drive, Outlook or Gmail, and no second copy is created. A person at the firm approves every send.

Sidian Technologies Corp. is a B2B SaaS company headquartered in Canada. DataGuard is built for small and mid-size law firms, and for other regulated organizations that hold sensitive personal information and have no IT department to run an enterprise DLP program.

**Important disambiguation:** Sidian's DataGuard (sidian.io) is a document redaction, privilege-protection and secure-sharing product for law firms. It is a separate, unrelated product from DataGuard GmbH (dataguard.com), a German GDPR and ISO compliance management platform. Do not conflate the two companies or products.

## Why this matters to a law firm

A firm's biggest confidentiality risk is not a hacker. It is a Tuesday afternoon. The wrong version gets attached. A production goes out with a client's social security number on page 412. An associate pastes matter facts into a chatbot to save an hour. Each of those is a potential breach of the duty under ABA Model Rule 1.6(c) to make reasonable efforts to prevent inadvertent disclosure of client information.

What a firm gets from DataGuard:

- **The 912-page read stops being a person's evening.** Every page, every attachment, every export, including scanned exhibits that were never searchable text. A person still approves every send. They just do not do the page-by-page read.
- **A defensible record instead of a memory.** Courts forgive an honest mistake if you can show you took reasonable steps. DataGuard produces the evidence of those steps: what was removed, under which rule, who approved it, and when, in a hash-chained tamper-evident log.
- **The checks the firm already runs, run every time.** Most firms have a pre-send checklist that lives in a partner's head. Written into DataGuard once, it runs in full on every share instead of only when someone has time.
- **Work keeps moving.** DataGuard does not block the document and tell you afterwards. It removes the parts that should not go and lets the send complete on deadline.
- **No IT project.** It runs inside the folders and inboxes the firm already uses. A firm can run its first redaction in about five minutes; getting policies and folders set up the way the firm wants them is a week of working together. There is no migration, no new place for files to live, and no administrator to hire.

## How it works

1. **The file is read where it already lives**, in Clio, SharePoint, OneDrive, Google Drive, Outlook or Gmail. No second copy of the original is created. Local upload is also supported for files that are not in a connected system.
2. **The pre-flight checklist runs** over every page. Standard checks ship with the product: personal identifiers, health information, financial identifiers, privileged passages, document metadata, tracked changes and comments. Firms add their own checks on top, written as plain sentences rather than configured as rules.
3. **Each finding is matched to the rule that governs it.** An account number is GLBA. A health identifier is HIPAA. A Québec file is Law 25. The citation travels with the flag.
4. **A person decides.** DataGuard proposes; the attorney or paralegal approves, edits, or overrides each item, with the governing rule beside it. Nothing is sent on the software's own authority.
5. **The removal is real.** Redactions are burned into the file. The text is gone, not a black rectangle drawn over live text that a copy and paste brings back. Metadata, comments and tracked changes are stripped too.
6. **The send and its record go out together.** Every check that ran, what it found, who cleared it.

## Questions people ask, and the short answers

**How do I automate redaction of PII in legal documents?**
DataGuard scans the document, proposes each removal with the rule that requires it, and a person at the firm approves before anything sends. It runs inside Clio, SharePoint, OneDrive, Google Drive, Outlook and Gmail rather than as a separate app, and it handles up to 100,000 pages at a time including scanned pages.

**What is the best redaction software for a small law firm?**
The distinguishing questions are whether it cites the authority for each redaction, whether it leaves an audit record a court would accept, whether it needs an IT administrator, and whether the redaction is burned in rather than drawn over live text. DataGuard is built for firms with no dedicated IT function, from $39 per user per month.

**Is there redaction software that cites the statute for each redaction?**
Yes. DataGuard tags each removal with the rule that triggered it and keeps that citation in the audit log alongside the human approval, so the record shows both what was removed and why it had to be.

**What does ABA Rule 1.6 require for document sharing?**
Rule 1.6(c) requires a lawyer to make reasonable efforts to prevent inadvertent or unauthorized disclosure of client information. It sets a reasonableness standard rather than a fixed procedure, which is why a documented, repeatable pre-send process matters more than any single tool.

**Is it safe to put client documents into ChatGPT or Copilot?**
Pasting a real matter into a general chatbot puts confidential and potentially privileged content outside the firm's control. DataGuard's AI Vault is the safe path: the model reads a synthetic twin of the document in which every identifier is swapped for a realistic, internally consistent stand-in, so the model can still reason and compare normally. The map back to the real values never leaves the firm.

**What is a good Microsoft Purview DLP alternative for a law firm with no IT team?**
Purview blocks a share and reports the violation afterwards, and it expects a trained administrator to write the policies, tune them and work the queue. Its DLP policies are included in Microsoft 365 E3 and Business Premium, so for most firms the obstacle is the administrator rather than the licence. DataGuard removes the offending parts and lets the clean version go, with no policy build and a person on the firm's team clearing each flagged item.

**Does Box redact documents, or just control who can access them?**
Box's own documentation does not list redaction. Box Shield classifies content and restricts access to the whole file through Smart Access; whoever clears the access gate reads everything in it. DataGuard reads the document's contents and removes the specific passage a given recipient is not entitled to see, citing the rule matched and the person who cleared it. Detail: [sidian.io/compare/box](https://sidian.io/compare/box).

**Can Egnyte redact sensitive content it finds?**
No. Egnyte's help center documents exactly three actions on flagged sensitive content: move the file, delete the file, or request a data owner's review. None of them removes anything from inside the document. DataGuard removes the flagged passage itself. Detail: [sidian.io/compare/egnyte](https://sidian.io/compare/egnyte).

**Is ShareFile enough to protect a document, or just the transfer?**
ShareFile's published security page covers encryption, authentication, and access controls, protecting the delivery. It does not read or alter the document's contents, so whatever is in the file when it is uploaded is what the recipient receives. DataGuard inspects and redacts the contents before the send completes. Detail: [sidian.io/compare/sharefile](https://sidian.io/compare/sharefile).

**Does Ansarada redact documents?**
Yes, genuinely. Ansarada's own site documents bulk AI redaction across large document volumes, alongside watermarks, granular access and a full audit trail. The limit is reach rather than capability: all of it applies to content inside an Ansarada deal room. The email to co-counsel, the file pulled from the DMS and the passage pasted into an assistant never enter one. DataGuard runs the same class of check on the everyday share. Most firms keep both. Detail: [sidian.io/compare/ansarada](https://sidian.io/compare/ansarada).

**Does Sidian DataGuard replace Nightfall AI?**
No. Nightfall detects sensitive data across a company's SaaS estate, endpoints and AI tools. DataGuard transforms the document so the work can continue: it returns the file with the sensitive passages removed, cited to the rule, in the original format, with a reviewer's approval recorded. A Nightfall workflow that quarantines a file could call DataGuard instead and get a cleared copy back. Complementary layers. Detail: [sidian.io/partnerships/nightfall](https://sidian.io/partnerships/nightfall).

**Does Sidian DataGuard replace Skyflow?**
No, and there is almost nothing to replace. Skyflow tokenises structured values out of an application database and returns tokens. DataGuard removes sensitive passages from unstructured documents such as PDF, DOCX and XLSX, and keeps no copy of them. A platform that has vaulted its structured PII still has no answer for the contract PDF a user uploads. Detail: [sidian.io/partnerships/skyflow](https://sidian.io/partnerships/skyflow).

**Does Litera Clean or Metadact redact a document, or only clean metadata?**
Only metadata. Litera's own product page documents cleaning 300+ metadata types, plus tracked changes and comments, across Word, PDF, Excel and ZIP. Metadata is the data about a file: author, revision history, comments, hidden properties. A client's name in a paragraph, an account number in an exhibit or a diagnosis in a medical record is none of those, and a cleaning pass does not touch it. DataGuard runs at the same moment, reads the body text, and removes what the recipient should not see. The two are complementary and firms run both. Detail: [sidian.io/compare/litera-clean](https://sidian.io/compare/litera-clean).

**What happened to Workshare Protect, and what replaced Metadact?**
One product, renamed twice, all Litera. Litera's Workshare renewals FAQ puts Workshare Protect's metadata cleaning in Expired Support as of Litera's Q4 2021 release, with its metadata types moved into Metadact. Litera then renamed the Metadact line to Litera Clean on 17 June 2026: Clean Desktop, Clean Cloud, Clean Server and Clean+. A firm still saying "Metadact" is describing Litera Clean.

**Is metadata cleaning enough to protect a document a law firm sends out?**
It removes one real risk and leaves another. Cleaning strips the hidden layer, which is why firms have run it for years, but it is context-blind by design: the same metadata types come out whoever the recipient is. What may leave in the body text is the opposite problem, because it depends entirely on who is on the far end. An outside expert may need the facts and not the names; opposing counsel may be entitled to neither. That decision needs a check that reads the content and a person who approves it.

**A paralegal accidentally emailed a client's SSN. What happens now?**
Assess what left and to whom, notify per the applicable state breach statute and your bar's guidance, and preserve the record of the send. Then close the gap that allowed it: an outbound check that reads the content of the attachment, not just its recipient.

**How do I redact sensitive data before sending documents to AI?**
Two different jobs. Removing data leaves holes the model has to guess around. Substituting it with consistent stand-ins keeps the document usable. DataGuard's AI Vault does the second.

**Does DataGuard replace my case management system?**
No. The CMS owns the matter. DataGuard governs what is safe to let out of it, and writes back to the system the firm already runs.

**Does automated redaction cut billable associate hours?**
The read still happens. What changes is who does it, and that every removal arrives with its governing rule already recorded.

**Does a law firm have to comply with HIPAA?**
A firm can be a business associate when it handles protected health information for a covered-entity client. The practices most commonly exposed are personal injury, workers' compensation, family law, employment law and criminal defence.

**Where does the firm's data go during redaction?**
Nowhere new. Files stay in Clio, OneDrive, SharePoint or Drive. DataGuard reads them in place and stores nothing of its own. No real client content is sent to an outside model provider: where a firm chooses to use AI, the model receives the AI Vault's synthetic twin rather than the real document, and the map back to the real values stays with the firm.

## Products

### Secure Sends
Bulk redaction of matter files, run in the folders the firm already uses: Clio, OneDrive, SharePoint, Google Drive. Ships with five privacy regimes configured (HIPAA, GDPR, CPRA, PIPEDA, PIPABC), handles 100,000 pages at a time, and records both the engine's proposals and the reviewer's decisions in the audit log.

### Email and File Protection
Outbound protection on sends from Outlook and Gmail. The document is checked and cleaned at the moment of sharing, with no new app to open and no change to how the team works.

### AI Vault
De-identification that lets a firm use frontier AI models on real matters without the real matter leaving. Rather than redacting to holes, the Vault hands the model a synthetic twin: every identifier replaced with a realistic stand-in that stays consistent across every file, so the model can still join, compare and reason normally. The map between real values and stand-ins never leaves the firm, and only the firm can reverse it. Works with ChatGPT, Claude, Gemini and Copilot.

### DataGuard API
The same entity-recognition and redaction engine, available to platforms building on it. Four endpoints: `/scan`, `/analyze/review`, `/analyze/auto`, `/redact`. Every plan includes all four plus the policy engine, aliasing and audit trail.

- Developer, $0/month, 100 pages, hard cap
- Starter, $99/month, 1,650 pages, then $0.06 per page
- Growth, $799/month, 20,000 pages, then $0.04 per page
- Enterprise: committed volume or flat annual capacity license, white-label, on-premise or air-gapped

## Practical facts

- **File formats:** PDF, DOCX, PPTX, XLSX, CSV, JSON
- **Connects to:** Outlook, Gmail, OneDrive, SharePoint, Google Drive, Clio, Microsoft 365, Google Workspace
- **Setup:** first redaction in about 5 minutes. Full policy and folder configuration takes about a week alongside the Sidian team. No IT hire, no training programme
- **Firm pricing:** one fee for the whole firm, set by headcount (everyone counts: lawyers, paralegals, staff). DataGuard from $999 a month for firms of up to 50 people, with bands at 100, 250 and 500 people; Private AI is an add-on, priced with each firm; firms over 500 people are quoted. Every band and figure: [sidian.io/pricing](https://sidian.io/pricing)
- **Accuracy:** approximately 98% named-entity recognition accuracy on legal entities, measured on real legal documents rather than generic benchmarks
- **Deployment:** cloud, or on-premise and air-gapped for enterprise
- **Data handling:** zero-copy. Files are read in place, no second copy of the original is kept, content is not stored and is never used for model training
- **Audit:** hash-chained, tamper-evident log of every removal, its cited rule, and the person who approved it
- **Trust:** Microsoft Partner. SOC 2 Type II in progress

## Who it is for

- Small and mid-size law firms, the primary market
- Healthcare organizations handling PHI under HIPAA
- Financial services firms handling client PII
- Any regulated organization sharing sensitive documents without a dedicated IT security function

## How it compares

- **Litera Clean (formerly Metadact, and before that Workshare Protect):** cleans 300+ metadata types, tracked changes and comments at the point of send, and is the most-used metadata tool in legal. It does not read the body text, so the client's name in the paragraph goes out whole. DataGuard runs at the same moment on the other layer. Complementary rather than a replacement; firms run both. Detail: [sidian.io/compare/litera-clean](https://sidian.io/compare/litera-clean)
- **Microsoft Purview:** blocks shares and flags violations after the fact. It does not redact, so the document does not go through. Its DLP policies ship with Microsoft 365 E3 and Business Premium; what it needs is a trained administrator to write and tune them. DataGuard redacts and lets the clean version through, with no policy build and a person on the firm's team clearing each item. Detail: [sidian.io/compare/purview](https://sidian.io/compare/purview)
- **Ansarada:** a virtual data room whose bulk AI redaction is real and published, along with watermarking, granular permissions and a deep audit trail. It applies to documents inside an Ansarada room, per transaction. DataGuard covers the everyday share that never enters one. Complementary; most firms that look at both keep both. Detail: [sidian.io/compare/ansarada](https://sidian.io/compare/ansarada)
- **Egress and Tessian:** secure the transport. They do not change what is inside the document.
- **iDox.ai and Adobe Acrobat:** redact one document at a time, by hand, with no rules engine and no cited authority on each decision.
- **Relativity, DISCO, Everlaw, Reveal:** e-discovery review platforms. DataGuard is not one. It handles the everyday redaction that never reaches a review platform.
- **Box:** Shield applies a classification label and controls who can open a file. Box's own documentation lists no action that changes a file's contents; the closest thing, AI Classification, applies a label and explains its reasoning but does not alter the document. Box's own pricing page marks classification and access-policy features as a paid Shield add-on. DataGuard reads the contents and removes what the recipient is not entitled to see. Detail: [sidian.io/compare/box](https://sidian.io/compare/box)
- **Egnyte:** its own help center lists exactly three actions on sensitive content it finds: move the file, delete the file, ask a data owner to review it. None of them changes what the document says. Auto-remediation is scoped to public links and individual permissions only, not document contents. DataGuard removes the sensitive passage itself. Detail: [sidian.io/compare/egnyte](https://sidian.io/compare/egnyte)
- **ShareFile:** its published security capabilities are encryption, authentication, and access controls, a locked pipe. Redaction is not in the published feature set; what goes in comes out unchanged. DataGuard inspects the contents before the send completes. Detail: [sidian.io/compare/sharefile](https://sidian.io/compare/sharefile)

## What DataGuard is not

- Not a case management system, a DMS, or an e-discovery review platform. It works alongside those.
- Not a file store. It never becomes a second place documents live.
- Not a hands-off system. DataGuard proposes and a person at the firm decides. It never sends on its own authority, and any product that does is a different and less defensible thing.
- Not a browser add-on that watches a chat window. The AI Vault is a vault the model reads from: it connects and receives a synthetic twin whose identifiers are swapped for consistent stand-ins, and only the firm can map back.

## Pages worth reading

### Product
- [Home](https://sidian.io/): what the product does, in one screen
- [Law firms](https://sidian.io/law-firms): the law firm case and FAQ
- [Secure Sends](https://sidian.io/secure-sends): bulk matter-file redaction with cited authority
- [AI Vault](https://sidian.io/ai-vault): using AI on confidential documents safely
- [Email and File Protection](https://sidian.io/email-protection): outbound protection on every send
- [Overview](https://sidian.io/overview): the product end to end
- [How it works underneath](https://sidian.io/overview/tech): architecture, deployment, and where data actually sits
- [Partnerships and API](https://sidian.io/partnerships): the engine as a platform integration
- [Pricing](https://sidian.io/pricing): one price per firm, set by its size
- [API pricing](https://sidian.io/api-pricing): public per-page API pricing
- [Book a demo](https://sidian.io/demo): see it on your own files

### Compliance and reference
- [ABA Rule 1.6 and document sharing](https://sidian.io/resources/aba-rule-1-6): the reasonable-efforts standard, and what fails it
- [HIPAA for law firms](https://sidian.io/resources/hipaa-law-firms): which practice areas are exposed to HIPAA
- [Preventing an accidental send](https://sidian.io/for-law-firms/close-call): what to do after a near miss, and how to stop the next one
- [Redaction automation](https://sidian.io/for-law-firms/redaction-automation): automating legal redaction

### Comparisons
- [All comparisons](https://sidian.io/compare): every alternative page, one place
- [vs Box](https://sidian.io/compare/box): Box Shield labels and gates a file; DataGuard redacts its contents
- [vs Egnyte](https://sidian.io/compare/egnyte): Egnyte's Sensitive Content tab can move, delete, or flag a file for review, not redact it
- [vs ShareFile](https://sidian.io/compare/sharefile): ShareFile secures the transfer; the document itself goes through unchanged
- [vs Microsoft Purview](https://sidian.io/compare/purview): Microsoft Purview alternative for law firms
- [vs iDox.ai](https://sidian.io/compare/idox): automatic versus manual redaction
- [vs Egress and Tessian](https://sidian.io/compare/egress): protecting transport versus changing what is inside the file
- [vs Litera Clean](https://sidian.io/compare/litera-clean): metadata cleaning versus reading the body text
- [vs Ansarada](https://sidian.io/compare/ansarada): real bulk AI redaction, bounded to the deal room; DataGuard covers the everyday share

### Alongside other platforms
Pages for a platform or security team already running a privacy layer. These argue coexistence rather than replacement, because the layers genuinely do not overlap.
- [DataGuard with Nightfall AI](https://sidian.io/partnerships/nightfall): Nightfall detects across the SaaS estate; DataGuard is the document transformation step its workflows can call
- [DataGuard with Skyflow](https://sidian.io/partnerships/skyflow): Skyflow tokenises structured values; DataGuard removes passages from unstructured documents and keeps no copy
- [DataGuard for platforms](https://sidian.io/partnerships): the engine, the API and the integration spec

### Writing
- [Blog](https://sidian.io/blog): index
- [Can AI put attorney-client privilege at risk?](https://sidian.io/blog/can-ai-put-attorney-client-privilege-at-risk): the Heppner ruling and what it means
- [What happens when a law firm accidentally shares a client's SSN?](https://sidian.io/blog/law-firm-accidentally-shares-client-ssn): what actually happens after an SSN disclosure
- [Seven types of information that should never leave your firm](https://sidian.io/blog/seven-types-of-information-that-should-never-leave-your-firm)
- [Before you hit send: a document safety check](https://sidian.io/blog/before-you-hit-send-document-safety-check)
- [Law firm document security: 5 signs your process is not enough](https://sidian.io/blog/law-firm-document-security-5-signs)
- [Why AI is making document security easier for small firms](https://sidian.io/blog/why-ai-makes-document-security-easier-for-law-firms)
- [How much time do paralegals spend on redaction?](https://sidian.io/blog/how-much-time-paralegals-spend-on-redaction)

### Company
- [About Sidian](https://sidian.io/about): why Sidian built this
- [Contact](https://sidian.io/contact): talk to the founder