PrivacyFirewall is an open-source, local-first data loss prevention tool that protects against accidental leaks of sensitive information when using AI chat services like ChatGPT, Claude, and Gemini. It operates entirely on your machine through a Chrome extension with two detection modes: fast regex-based pattern matching for secrets (API keys, emails, phone numbers, credit cards) and an optional local transformer model for deeper PII detection (names, organizations, locations). The system intercepts paste events and warns during typing without sending any data to external servers, combining browser-based detection with an optional local FastAPI service running a BERT NER model.

โ€ข5m read timeโ€ขFrom github.com
Post cover image
Table of contents
PrivacyFirewall (Local AI Privacy Shield / Local LLM DLP)๐Ÿš€ Quickstart (Local Development)๐Ÿ” Detection Coverage๐Ÿ“ Project Layout๐Ÿงช Development๐Ÿ› ๏ธ Future Improvementsโ— Troubleshooting๐Ÿค Contributing๐Ÿ” Security & Privacy Notes๐Ÿ“„ License

Sort: