SafePrompt 作者： SafePrompt

SafePrompt protects your sensitive data before it reaches ChatGPT, Claude, Gemini, DeepSeek, Qwen, OpenRouter, and Mistral Le Chat.

100% LOCAL — NO DATA LEAVES YOUR BROWSER

All processing (detection, tokenization, rehydration) happens inside your browser. SafePrompt never sends your prompts to a third-party server. GDPR-compliant by design.

AUTOMATIC DETECTION AND MASKING

Before your prompt is sent, SafePrompt detects:
- Secrets: API keys (AWS, GitHub, Stripe, OpenAI), JWT tokens, plaintext passwords
- Personal identifiers: email, phone, IBAN, credit card number
- French identifiers: SIRET, SIREN, social security number (NIR), VAT number
- Postal addresses, dates of birth

Secrets are blocked (fail-closed); personal data is automatically replaced with tokens (⟦EMAIL_1⟧, ⟦FR_SIREN_2⟧, etc.) before sending.

TRANSPARENT REHYDRATION

When the LLM replies with the tokens, SafePrompt restores the original values — only inside your browser. Hover a restored token → tooltip confirms "SafePrompt restored this value locally". The LLM never saw the original data.

7 SUPPORTED PLATFORMS

• ChatGPT (chatgpt.com)
• Claude (claude.ai)
• Gemini (gemini.google.com)
• Qwen (chat.qwen.ai)
• DeepSeek (chat.deepseek.com)
• OpenRouter (openrouter.ai)
• Mistral Le Chat (chat.mistral.ai)

TEAM DASHBOARD

Solo Paid, Starter, and Business plans include a centralized dashboard on safeprompt.m-kis.fr: incident visibility, seat management, audit log, per-tenant anonymization (SHA256 + salt), GDPR export.

PLANS

• Solo Free — free, secrets detection only
• Solo Paid — €5/month, full secrets + PII detection
• Starter — €12/seat/month (5–49 seats), team dashboard
• Business — €8/seat/month (50+ seats), anonymization, priority support

SECURITY AND COMPLIANCE

• No "<all_urls>" permission — host_permissions limited to the 7 supported LLMs
• GDPR by default: no personal data sent to third parties
• Manifest V3 compliant, regular dependency audits

SafePrompt is published by M-KIS (https://m-kis.fr), a Lorraine-based agency specialized in cybersecurity and responsible AI.

Support & questions: contact@m-kis.fr