Protect Personal Data with Open-Source Proxy for Secure AI Interactions

Post Views: 3

Open-source privacy proxy safeguards sensitive data before it reaches external AI services

In recent years, businesses have increasingly relied on artificial intelligence solutions to streamline operations and enhance decision-making processes.

However, this increased reliance on AI also raises concerns about data protection, particularly when sensitive customer information is transmitted to external AI services.

To mitigate these risks, Dataiku has developed an open-source solution called Kiji Privacy Proxy, which effectively detects and conceals personally identifiable information (PII) prior to transmission.

Kiji Privacy Proxy: How it Works

Local Gateway:

Kiji Privacy Proxy operates as a local gateway between internal applications and external AI APIs.
Machine Learning Model:

It utilizes a sophisticated machine learning model designed to identify various types of PII, including addresses, phone numbers, Social Security numbers, credit card numbers, and IP addresses.
Data Concealment:

Once sensitive details are identified, they are replaced with realistic placeholder values, allowing the modified request to proceed while protecting the underlying data.
Low Latency:

The detection process occurs locally on the user’s device, utilizing a quantized DistilBERT model executed via ONNX Runtime, eliminating the need for any external service calls during the detection phase.

According to the project documentation, response times remain below 100 milliseconds for most requests.

Distribution and Platforms

macOS:

Users can install a native Electron desktop application, which configures Proxy Auto-Config to direct traffic from Safari and Chrome through Kiji on port 8081 without requiring manual adjustments.
Linux:

A standalone server binary is available, along with the ability to set HTTP_PROXY and HTTPS_PROXY values.
Chrome Extension:

A dedicated Chrome extension facilitates seamless integration for users interacting with services like ChatGPT through their browsers.

Motivation and Compliance Requirements

The motivation behind Kiji Privacy Proxy lies in addressing compliance requirements tied to the transfer of sensitive data.

Regulators’ expectations under GDPR, HIPAA, and CCPA necessitate careful control over data leaving organizational perimeters.

According to a 2026 Dataiku survey, a staggering 85% of surveyed CIOs acknowledged seeing AI-related projects either delayed or blocked due to shortcomings in traceability or explainability, with privacy concerns being a primary reason.

Conclusion

Dataiku offers Kiji Privacy Proxy for free on GitHub, aiming to empower organizations in safeguarding sensitive information as they leverage AI solutions.