Your prompts,
never exposed.
Blur redacts sensitive data before it reaches ChatGPT, Claude, or Gemini. Detection runs entirely in your browser. Under 10ms.
I'll help you draft that clinical update. What specific information would you like to include?
Compatibility
Works wherever your team uses AI.
How it works
One prompt, protected end to end
Follow a real prompt through Blur — from the first keystroke to what the AI actually receives.
Type like you always do
Blur runs inside ChatGPT, Claude, and Gemini. As you type, it scans your prompt locally and the floating pill keeps a live privacy score — nothing leaves your browser.
Draft an update for Sarah Johnson at sarah@google.com about patient progress
Click the pill to review
Every detection is listed with its severity — here Blur caught a patient’s name and email. Financial IDs, health details, and a dozen more categories are covered the same way.
Review and protect your information before sending.
Mask everything in one click
One click swaps real values for safe placeholders that keep your prompt’s meaning intact. Prefer precision? Mask items individually instead.
Draft an update for [name] at [email] about patient progress
Send with confidence
The model only ever sees placeholders. The original values stay on your device — Blur never stores or transmits them.
Draft an update for [name] at [email] about patient progress
Detection
Four categories of sensitive data, caught in real time
SSNs, credit cards, PHI, contact data, names, and 20+ other PII patterns. Aggressive Mode is recommended for HIPAA, PCI, and SOX-adjacent workflows.
Financial & IDs
SSN: 123-45-6789Card: 4111-****-****-1234Contact & Addresses
sarah@example.com+1 (415) 555-0123Names & Entities
Sarah JohnsonSutter HealthSensitive Attributes
Type 2 DiabetesDOB: 08/12/1984Built for regulated teams
Used by teams that can't afford prompt leaks.
Blur is designed for organizations where a single exposed identifier creates real regulatory or liability risk.
Legal
Keep client names, matter details, and privileged context out of model traffic. Detection runs locally before the prompt reaches any LLM.
Healthcare
Mask PHI including names, dates of birth, and health terms before a prompt is sent. Aggressive Mode is recommended for HIPAA-adjacent workflows.
Government & regulated
Deploy through Chrome Enterprise with ExtensionInstallForcelist. Zero-data architecture keeps prompt content on the user’s device end-to-end.
"ExtensionInstallForcelist": [
"<blur-extension-id>"
]Team Management
Deploy to your whole org, from one dashboard.
Force-install through Chrome Enterprise, invite teammates by email, and let seat-based billing prorate itself as the team changes.
Pricing
Predictable seats, billed when teammates join.
Blur charges only for active members. Seats are prorated automatically and synced to your Stripe subscription whenever a teammate accepts an invite.
Free
Free plan for everyone. Create an account and mask names, emails, and phone numbers on every supported AI site.
- Masks people's names, email addresses, and phone numbers
- ChatGPT, Claude & Gemini compatibility
- Settings sync across every supported site
- Free account, no credit card required
Individual
Solo paid plan for power users. All detection, allowlist, and personal analytics.
- Every detection type across all 4 categories
- Adds locations, links, companies, schools, Financial & IDs, Sensitive Attributes
- Aggressive Mode (also flags dates, IDs, and rare patterns)
- Smart Placeholders with realistic replacement names
- Personal allowlist — skip detection on your own terms
- Personal analytics: redactions, categories, top rules
Enterprise
Team plan with admin dashboard, shared rules, and per-member analytics.
- Everything in Individual, per seat
- Shared admin dashboard (team, billing, policies)
- Role-based access control (owner, admin, member)
- Shared custom rules, allowlist, detection policies
- Per-member analytics and alerts
- Priority customer support
FAQ
Common questions
Keep your team's prompts private.
Deploy Blur in minutes. No code changes, no workflow disruption.