PrivyPrompt - Privacy Firewall for AI

PrivyPrompt is a local-first privacy firewall and prompt security extension for supported AI writing, coding, research, and chat tools. Before you send a prompt or upload a supported file, PrivyPrompt scans locally in your browser and helps you review, protect, or block sensitive information. It is built for users who want safer AI workflows, prompt privacy, AI prompt security, API key protection, document privacy, and prompt-injection awareness without sending their content to another cloud service. PrivyPrompt can detect personal details, national identifiers, financial data, developer credentials, healthcare identifiers, confidential business terms, and risky AI-tool instructions such as prompt-injection, jailbreak, system-prompt leakage, local-secret access, and secret-exfiltration wording. Key features: - Local prompt scanning on supported AI websites - Detection for API keys, credentials, personal information, financial data, healthcare identifiers, and confidential business content - Local AI tool security warnings for prompt-injection, jailbreak, system-prompt leakage, local-secret access, and secret-exfiltration patterns - 47 dedicated country packs with checksum and structural validation where reliable public validation rules are available - Local PDF and DOCX text extraction - Offline English OCR for clear PNG, JPG, WebP, and BMP images - Per-finding review with Allow, Warn, Protect, and Block policies - Reversible session-only aliases that keep original values out of AI prompts - Secure local AI-response restoration without inserting originals into the AI website's readable page DOM - Local GDPR, HIPAA, PCI, and developer-secret policy presets; these shortcuts are not legal or regulatory certification - No PrivyPrompt account, advertising, telemetry, or uploads to PrivyPrompt Why users install PrivyPrompt: - Protect API keys, tokens, passwords, private keys, and developer secrets - Review sensitive data before sending prompts to AI tools - Reduce accidental exposure of personal, financial, healthcare, and business data - Get local warnings for prompt-injection and jailbreak-style text - Scan supported files such as PDF, DOCX, text, code, configuration, CSV, JSON, Markdown, logs, and clear images - Use reversible session-only aliases instead of sending original sensitive values PrivyPrompt processes supported prompt and file content inside your browser before submission. The destination AI service receives the content that the user approves and sends. Settings and aggregate counters are stored locally. Complete prompts, files, detected values, and private alias mappings are not transmitted to the developer. PrivyPrompt is a privacy aid and does not provide legal or regulatory certification. Always review protected content before submission.