Product tour

One extension, from detection to audit.

InputGuard runs in the browser on the AI sites you protect. Follow this browser-native AI DLP flow end to end — what it detects, how it decides, how you deploy it, and what the audit trail keeps.

Join early access See the security model

Free for individuals • per-device pricing for organizations

01Detect 02Decide 03Deploy 04Audit

Detect

Find the data that actually causes problems.

Detection runs on the device against three categories, at every point data can leave an AI workflow. You decide which categories matter on each site.

Messagescanning locally…

Deploy failed again — here's the prod key AKIA…EXAMPLE. Ping me at jordan@acme.com or (415) 555-0148 once the Project Atlas migration is unblocked.

4 findings1 credential2 personal1 business term

Matched on the device — nothing was uploaded to scan it.

Credentials & secrets
API keys, access tokens, private keys, and password-like strings.
Personal data (PII)
SSNs, payment card numbers, emails, phone numbers, and bank details.
Business-sensitive terms
Your own keywords — project names, customer identifiers, contract language.

On paste

Scanned the instant it lands in an AI tool — before it becomes part of a prompt.

On submit

Final prompt text is checked at send time, with the person in the loop.

On attachment

Supported text files (TXT, CSV, JSON, Markdown, logs) are read in the extension.

Decide

One detection, four ways to respond.

Policy decides how InputGuard reacts — tuned per site, from a light review to a hard block. Here's exactly what each level looks like to the person at the keyboard.

Review

InputGuard surfaces exactly what was detected and lets the person decide in the moment — send as-is, send redacted, or cancel. They stay aware of the risk without losing their flow.

When to use Lowest-friction enforcement, for sites and categories where awareness is enough and you trust people to make the call.

Data categories

Switch on the categories that matter.

Every response level draws from one policy. Turn categories on or off and set how strict each one is — then open the full editor to fine-tune.

Identity

SSN
Email
Phone

Financial

Credit cards
Bank accounts

Secrets & credentials

API keys
Passwords
Private keys

Business sensitive

Contracts
Customer lists
Internal terms

Deploy

Roll it out the way your team works.

The same extension runs for one person or a whole fleet. One panel governs browser-native DLP behavior; clear states decide which AI sites it touches.

Protection on or off
Turn scanning and blocked-site enforcement on or off for everyone on the policy.
Paste protection
Scan pasted text before it ever enters an AI tool.
Prompt scan
Check the final prompt locally when you send — prompt text is never stored or uploaded.
Text file attachments
Read and scan supported files — TXT, CSV, JSON, Markdown, and logs — before submit.
Blocked-site behavior
Hard-block AI sites that policy does not allow.

AI site controls

Decide how each AI site is treated.

Sort the AI tools your team uses into clear states. Data Loss Prevention for AI tools applies where you want it, and new sites get a safe default instead of slipping through.