AI privacy firewall

Let your team use AI without leaking sensitive data.

PromptShield is a local-first protection layer for ChatGPT, Claude, Gemini, Copilot and other AI tools. It detects risky data before sending, creates a safe prompt and helps teams keep AI use under control.

Try scanner demo View platform plans

Private use €0 Team plans per user Runs locally first Redact · Warn · Block

Before PromptShield Risk detected

Draft a reply to Anna Bērziņa. Include her personal code 010180-12345, email anna@example.com and the confidential HR settlement summary.

Local scan Safe version created

After PromptShield Ready to use

Draft a reply to [PERSON_NAME_REMOVED]. Do not include [PERSONAL_CODE_REMOVED], [EMAIL_REMOVED] or [CONFIDENTIAL_CONTEXT].

Built for companies where employees use AI with real business information.

Professional firms IT & SaaS teams HR & recruitment Finance & brokers Healthcare-adjacent teams Support departments

The risk

AI use is already happening. Most companies do not control what gets pasted.

Employees use AI for emails, translations, summaries, coding, HR cases, contracts and client questions. The problem is not only the AI provider — it is the moment sensitive data is copied into a prompt.

Personal data

Names, personal codes, emails, phone numbers, passports, addresses and employee data can be pasted into public AI tools.

Company secrets

Source code, API keys, access tokens, product documentation, client files and internal issues can leave the company unnoticed.

No practical control

A written AI policy is useful, but employees still need a simple tool that warns or redacts before the risky prompt is sent.

The platform

A privacy layer between employees and AI tools.

PromptShield starts with a browser extension and grows into a team platform for safe AI use: local scanning, redaction, team rules, usage visibility and admin controls.

Scan locally

Detect sensitive patterns before text is submitted to AI tools.

Show risk

Highlight what was found and explain why it may be unsafe.

Create safe prompt

Replace sensitive data with placeholders while keeping the prompt useful.

Apply team rules

For companies: allow, warn, redact or block based on admin settings.

Product modules

Start simple. Add company controls when the team grows.

Browser extension

Protects prompts directly where employees use AI websites.

Local redaction

Creates a safer version before sensitive values leave the device.

Team dashboard

Company view for users, rules, detected categories and basic usage signals.

Admin rules

Choose what should be allowed, warned, redacted or blocked for the team.

Interactive demo

Paste a prompt. See the risk. Create a safe version.

This demo runs in the browser and does not send your test text anywhere. It shows the first product experience: local detection and redaction before using an external AI tool.

Names in context, email addresses and IBANs
Latvian and Finnish personal identifiers
Card-like numbers, API keys, passwords and tokens
Confidentiality, HR, legal and medical keywords

Test prompt

The scanner runs locally in your browser. Try editing the prompt or click the buttons again.

Safe prompt

Platform pricing concept

Simple subscription pricing for private users and teams.

Pricing is shown for validation. The first version should be easy to understand: free private use, paid team control per user.

Private

€0

For individuals who want to check prompts before using public AI tools.

Local prompt scanner
Safe prompt generation
No company dashboard

Team

€9/user/month

For small teams that want safer AI use without enterprise DLP complexity.

Browser extension
Shared team rules
Warn, redact or block

Business

€19/user/month

For companies that need central settings and visibility across users.

Admin console
Usage and risk categories
Metadata logs without raw prompts

Optional onboarding, AI policy and employee training can be offered separately. The platform subscription stays focused on the tool.

Early access

Get access to the first private version.

We are preparing the first PromptShield prototype for private users and selected teams. Join the early access list if you want to test the scanner, browser extension concept or team controls.