ProxyDLP is an open-source monitoring proxy designed for security engineers in enterprise environments. It enables secure, auditable use of AI tools like ChatGPT while detecting and preventing potential data leaks through user conversations and file uploads.
Its allows enterprise users to access AI tools like ChatGPT while ensuring visibility, data security, and compliance. It enables organizations to benefit from AI productivity tools without sacrificing confidentiality or control over sensitive information.
⚠️ This is not a privacy or anonymization tool — ProxyDLP is built to observe and control AI tool usage across an organization.
ProxyDLP helps organizations:
- Monitor and inspect conversations with AI assistants
- Detect confidential or sensitive data in uploads and messages
- Link AI usage to individual users or accounts
- Provide a centralized interface for reviewing activity and enforcing policy
- Ideal as a technical tool to provide compliance for ISO/IEC 27001 and ISO/IEC 42001
- Proxy-based inspection — AI-related traffic is routed through a local MiTM proxy (port
8080) - Conversation monitoring — Intercept and inspect chat requests and responses
- File inspection — Decode, extract, and analyze contents of uploaded PDFs, Excel files, and images (OCR)
- Pattern-based detection — Configurable regular expressions detect potential data leaks
- Semantic topic matching — Discover and match topics in conversations and files using Faiss vector indexes for efficient similarity search
- Dashboard interface — Real-time view of events, traffic, and alerts for security engineers (available on HTTP
80and HTTPS443)
- Docker + docker-compose
- (Optional) Custom TLS/CA certificates
# 1. Generate certificates and secrets
./generate_secrets.sh
# 2. Launch services
docker-compose up| Component | Port | Description |
|---|---|---|
| ProxyDLP UI | 443 | Secure web interface for monitoring (HTTPS) |
| 80 | Web interface (HTTP fallback) | |
| Monitoring Proxy | 8080 | MiTM proxy for AI traffic |
Make sure client traffic to AI tools is routed through the proxy, e.g., via system proxy settings or PAC files.
- The proxy intercepts TLS traffic using a custom Root CA (
mitmCA.pem) - You must configure your client machines to trust this CA certificate (for example via GPO deployment).
- The proxy inspects and decodes traffic from supported AI platforms
- All activity is logged and linked to the user or session that initiated it
- Default user is admin and password is admin (can be changed after installation)
- It is recommended to configure a PAC file on the client machines (i.e. via GPO) so that only the traffic intended to the AI tools goes through the proxy, excluding the rest of the traffic. The PAC file can be automatically generated from the "SITES" page.
ProxyDLP is aimed at:
- Security engineers
- SOC analysts
- IT compliance teams
It is intended for use inside organizations that wish to embrace AI tools without sacrificing security oversight.
We welcome community contributions!
- Fork the repository
- Create a feature branch:
git checkout -b my-feature - Make your changes
- Commit and push:
git commit -m "Add feature"→git push - Open a Pull Request
We recommend keeping PRs focused and writing meaningful commit messages.
- Add support for more AI tools
- Improve PDF, Excel, and OCR parsing
- Enhance the dashboard UI/UX
- Add log filtering, alerting, or export options
If you discover a security vulnerability, please report it privately:
- Contact: f.amez1992@gmail.com
- Do not create public issues for security-related matters
- We follow responsible disclosure best practices