Privacy & Data Control

WebLLM puts you in control of your data and privacy. Unlike traditional AI integrations where your data flows through the website’s servers, WebLLM lets you choose exactly how your data is handled.

Your Data, Your Choice

With WebLLM, you have three privacy levels to choose from:

🔒 Maximum Privacy: Local Models

Your data never leaves your device.

When you use local models:

• All AI processing happens on your computer
• Zero network requests for AI processing
• Works completely offline
• No one can see your data - not websites, not AI companies, not even us

Perfect for:

• Sensitive documents
• Personal information
• Private conversations
• Offline work
• Maximum privacy

How it works:

Your Data → Local Model (in your browser) → Response
          ↑
    Never leaves your computer

🔐 High Privacy: Your Own API Keys

Your data goes directly to the provider you choose.

When you use your own API keys:

• Data goes directly from the extension to your chosen provider
• The website never sees your data
• You control which provider processes your data
• You can delete your data from the provider anytime

Perfect for:

• Premium AI capabilities
• When you trust specific providers
• When you want to use existing subscriptions

How it works:

Your Data → WebLLM Extension → Your Chosen Provider
          ↑
    Website never sees it

⚙️ Hybrid: Best of Both Worlds

Smart routing for optimal privacy and performance.

Set up multiple providers and let WebLLM automatically:

• Use local models for quick, simple tasks
• Use cloud providers only for complex tasks
• Minimize data sharing while maintaining great performance

What Makes WebLLM Private?

Traditional AI (How Most Websites Work)

You → Website's Server → AI Company's Server
     ↑                 ↑
   Website sees     AI company sees
   your data        your data

Problems:

• Your data passes through two parties
• You don’t know how long they keep it
• You can’t easily delete it
• You have no choice in providers

WebLLM

You → WebLLM Extension → [Your Choice:]
                         ├─ Local Model (data stays)
                         ├─ Claude (your API key)
                         └─ GPT-4 (your API key)
     ↑
   Website never touches your data

Benefits:

• Your data only goes where you send it
• You can delete everything at any time
• You choose who processes your data
• Full transparency - you can see every request

Data Retention Control

You decide how long WebLLM keeps request history:

Retention Options

• Don’t save anything - No history kept (maximum privacy)
• 24 hours - For temporary debugging
• 7 days - For recent history review
• 30 days - For monthly analysis
• Forever - Until you manually delete

Change anytime in Settings → Privacy → Data Retention

What Gets Saved (Locally Only)

When you enable history, WebLLM saves on your computer only:

• Your prompts and AI responses
• Which provider was used
• Timestamp of requests
• Website that made the request

Never saved or shared:

• Your API keys (stored separately, encrypted)
• Your personal information
• Your browsing history
• Any analytics or tracking

Permission System

WebLLM asks your permission before any website can use AI:

First-Time Permission

When a website tries to use WebLLM:

1. A notification appears asking for permission
2. You can allow or deny
3. You can choose “Remember this decision”
4. You can revoke permission anytime

Per-Website Control

In the extension, you can see:

• Which websites have permission
• How many requests each has made
• When they last used AI
• Revoke access with one click

Privacy Features

🔐 Encrypted Storage

All data stored by WebLLM is encrypted:

• API keys encrypted with Chrome’s secure storage
• Local models stored in browser’s encrypted database
• Request history (if enabled) encrypted

🚫 No Tracking

WebLLM never:

• Sends analytics or telemetry
• Tracks your browsing
• Shares data with third parties
• Phones home to our servers

📊 Full Transparency

You can always see:

• Every AI request made
• Which website made it
• Which provider handled it
• The exact prompt and response

View in Extension → History

🗑️ Easy Deletion

Delete your data anytime:

• Per-request deletion - Remove individual requests
• Per-website deletion - Clear all data from one site
• Complete deletion - Clear all WebLLM data
• Export first - Download your data before deleting

Compliance Benefits

For You

WebLLM makes it easy to exercise your privacy rights:

Right to Know ✅ - See exactly what data exists Right to Delete ✅ - Delete everything with one click Right to Export ✅ - Download all your data Right to Control ✅ - Choose providers and retention

For Websites Using WebLLM

Websites using WebLLM have fewer compliance obligations:

• No PII on their servers - They never see your AI data
• No data processing agreements - You control processing
• Simpler privacy policies - They just enable the feature
• Reduced liability - They don’t handle sensitive data

Privacy Comparison

Feature	Traditional AI	WebLLM Local	WebLLM Cloud
Website sees your data	✅ Yes	❌ No	❌ No
AI company sees your data	✅ Yes	❌ No	⚠️ Your choice
You control provider	❌ No	✅ Yes	✅ Yes
Works offline	❌ No	✅ Yes	❌ No
Data retention control	❌ No	✅ Yes	✅ Yes
Instant deletion	❌ No	✅ Yes	✅ Yes*
Audit trail	❌ No	✅ Yes	✅ Yes

*Local history deleted instantly; cloud provider deletion per their policy

Privacy Best Practices

Maximum Privacy Setup

For the most private experience:

1. Use local models only

   • Enable local model provider
   • Disable cloud providers
   • Data never leaves your device

2. Disable history

   • Settings → Privacy → Data Retention → “Don’t save”
   • No local storage of requests

3. Review permissions regularly

   • Check which sites have access
   • Revoke unused permissions

Balanced Privacy Setup

For privacy with premium capabilities:

1. Use reputable cloud providers

   • Choose providers you trust (Anthropic, OpenAI)
   • Use your own API keys
   • Read their privacy policies

2. Set short retention

   • Keep 7-day history for debugging
   • Auto-delete old requests

3. Use local for sensitive tasks

   • Set local model as priority #1
   • Cloud providers as fallback
   • Sensitive data stays local

Questions About Privacy

Can websites steal my API keys?

No. Websites never have access to your API keys. WebLLM keeps them encrypted and separate.

Can WebLLM see my API keys?

The extension needs your API keys to call providers, but:

• They’re encrypted in Chrome’s secure storage
• Never sent to WebLLM servers (we don’t have servers!)
• Only used to make API calls you approve

What if a website is malicious?

Malicious websites cannot:

• Access your API keys
• See your history with other websites
• Make requests without permission
• Exceed rate limits you set

You always have to grant permission first.

Is local processing really private?

Yes! Local models:

• Run entirely in your browser using WebGPU/WebAssembly
• Make zero network requests
• Are fully auditable (open source)
• Can be used offline to prove no data is sent

Can I use WebLLM at work?

Check with your IT department, but WebLLM can be more compliant than traditional AI:

• Local models keep sensitive data on your machine
• You can use company-approved API keys
• Full audit trail of all AI usage
• No third-party services receiving data

How do I verify WebLLM’s privacy?

WebLLM is open source:

• Review the code on GitHub
• Community security audits
• Use browser developer tools to see network requests
• Try local model offline - confirm no network activity

Take Control of Your Privacy

Ready to use AI on your terms?

➡️ Install WebLLM to get started ➡️ Configure Providers to set up privacy options ➡️ Technical Privacy Details for developers

---

Your data. Your choice. Your control.