Lawyers can now use AI
without risking their
privilege or work product.

Privilege Protector strips your clients’ identities from every AI prompt, locally, before anything leaves your machine. Work product protection on every transmission. Compatible with Claude, GPT-4, Gemini, and 200+ models.

Start Your Free Trial See how it works ↓

Windows & Mac · Solo, Firm, Enterprise

Step 1 of 5

Write your prompts the way you always do

No change in how you work. Use real names, real facts, real case details.

New Prompt

RE: Motion to dismiss, Johnson matter

My client has been sued for breach of contract.

The incident occurred on March 14, 2024 at

742 Evergreen Terrace. Please advise.

Step 2 of 5

Sensitive information is quietly identified

Privilege Protector scans your prompt locally. Nothing has left your machine yet.

Scanning…

RE: Motion to dismiss, Johnson matter

My client has been sued for breach of contract.

The incident occurred on March 14, 2024 at

742 Evergreen Terrace. Please advise.

NameDateAddress

🔒 Identified on your machine. Not yet transmitted.

Step 3 of 5

Protected before anything leaves your computer

Sensitive details are secured locally. A work product header is added. Only then does anything transmit.

Your prompt
with real details

→

Privilege Protector

Your machine. Local only.

→

Protected
identities secured

Fed. R. Civ. P. 26(b)(3) work product header attached to every transmission

Step 4 of 5

Your AI model gets to work

Claude, GPT-4, Gemini, or any of 200+ models analyze your matter. Your client’s identity remains secure throughout.

Claude

GPT-4o

Gemini

+ 197 more

💡 Analyzing…

Based on the facts presented, the strongest grounds for dismissal under Rule 12(b)(6) are the statute of limitations and failure to state a claim. The incident date places the matter squarely within…|

Step 5 of 5

Complete, accurate, and privileged

Real names and details are restored locally before you see the response. Your client never left your machine.

✓ Response delivered. Privilege protected.

Based on the Johnson matter facts, the strongest grounds for dismissal are the statute of limitations. The March 14, 2024 incident date places this squarely within the limitations period. Failure to state a cognizable breach at 742 Evergreen Terrace…

📄 Session logged locally · 26(b)(3) asserted · 0 client details transmitted

The Heppner problem

AI is a privilege trap.
Most attorneys don’t know it yet.

When you send a client’s name to an AI service, you may be disclosing privileged information to a third party, waiving the very protection your client is depending on.

⚠ Controlling Authority

United States v. Heppner

The court found that transmitting privileged communications to an AI platform whose terms of service permit use of that data, constituted a disclosure to a third party, destroying the protection. The attorney did not intend to waive privilege. The court did not care.

D. Minn. · See also Hickman v. Taylor, 329 U.S. 495 (1947)

⚠

Every prompt is a potential disclosure

When client names, case numbers, and facts travel to Anthropic, OpenAI, or Google, you have transmitted them to a third party. The AI company’s data practices, not your intent, determine what happens next.

⚠

Competence requires understanding the risk

Model Rule 1.1 requires technological competence. Bar associations in California, New York, Florida, and elsewhere have issued guidance on AI and confidentiality. Ignorance is not a defense.

✓

There is now a documented remedy

Privilege Protector intercepts every prompt on your machine before transmission, strips identifying information, sends only anonymized text, and restores real names in the response. The AI never sees your client’s identity.

How it works

Three steps. Zero client data transmitted.

The entire anonymization process runs on your machine. By the time your prompt reaches any AI server, your client is gone.

You type your prompt normally

Write your prompt the way you always would. Use real names, real case numbers, real facts. Privilege Protector works in the background. You don’t change how you write.

PII protected locally before transmission

Our engine identifies names, addresses, case numbers, SSNs, dates, phone numbers, emails, dollar amounts, and more. Each is secured locally. A work product header is added. Only the protected version ever leaves your machine.

Real names restored in the response

The AI responds using the tokens. Privilege Protector intercepts that response locally, swaps every token back to the real value, and delivers the final answer with your client’s name restored.

What’s included

Everything you need.
Nothing you don’t.

Built for legal practice. Not adapted from a general purpose tool.

Work product header on every prompt

Every transmission is automatically wrapped in a Fed. R. Civ. P. 26(b)(3) assertion, Kovel notice, and California absolute work product declaration. Your firm name appears in every header.

🛡

Local First PII Engine

Names, addresses, case numbers, SSNs, dates, phone numbers, emails, dollar amounts, medical records, EINs, PO boxes, estate and trust names. All detected and replaced before anything leaves your machine.

📄

Document anonymization

Upload PDF, Word, and Excel files. Privilege Protector extracts text, strips PII, sends anonymized content to the AI, and restores real values in the response. Up to 10 documents per prompt.

🤖

Any AI model, one key

Claude Sonnet, GPT-4o, Gemini, Grok, Llama, and 200+ models via OpenRouter. Pick the right model for each task. One OpenRouter account covers everything.

Requires a separate OpenRouter API account. OpenRouter usage fees are billed directly by OpenRouter and are not included in your Privilege Protector subscription.

📊

File generation with PII restored

Ask the AI to create a spreadsheet or Word document. Privilege Protector generates the file with real names restored and ready to download immediately.

🔍

Audit log for every session

Every session is logged locally: timestamp, PII category count, disclaimer status. If a privilege challenge ever arrives, you have a documented record of the protection measures in place.

Supported AI models

Your practice. Your model.

You choose which AI handles your work. Privilege Protector protects your clients regardless of which model you select.

Claude Sonnet 4.6

Claude Opus 4.6

Claude Haiku 4.5

GPT-4o

GPT-4o Mini

Gemini 2.0 Flash

Gemini 2.5 Pro

Grok 3

Llama 3.3 70B

+ 190 more via OpenRouter

You supply your own OpenRouter API key. We never see your API credentials or prompt content.

Pricing

Simple, transparent pricing.

Per seat. No seat minimums on Firm tiers. 10-day free trial on every plan.

Monthly Annual Save 15%

Solo

1 attorney seat

$ 79

/seat/month

Billed $804/year

1 attorney seat maximum
All 200+ AI models via OpenRouter
Local PII engine — never transmitted
Work product header on every prompt
26(b)(3) assertion on every transmission
Browser extension (Chrome and Edge)
10-day free trial

Start 10-Day Free Trial

Credit card required. Cancel before day 10 at no charge.

Lawyers can now use AIwithout risking theirprivilege or work product.

AI is a privilege trap.Most attorneys don’t know it yet.