Dataiku Launches Open-Source Privacy Layer for Generative AI Data Protection
Kiji Privacy Proxy enables enterprises to harness external AI services without exposing personally identifiable information, removing a critical barrier to AI adoption
With enterprises racing to operationalize generative AI, a new and urgent question has emerged: How do you unlock its power without putting sensitive data at risk? Dataiku, the Platform for AI Success, announced the general availability of Kiji Privacy Proxy™, a new open-source privacy layer that ensures personally identifiable information (PII) never leaves an organization’s control, even when using third-party AI services such as OpenAI or Anthropic.
Read More: ITTech Pulse Exclusive Interview with Stanley R. Hughey, Chief Technology Officer at iNet
Kiji Privacy Proxy is the latest innovation from Dataiku’s 575 Lab, the company’s open-source initiative focused on advancing transparency, privacy, and responsible AI. As organizations scale AI, the risk of exposing sensitive data has become a critical barrier to adoption. Kiji Privacy Proxy addresses this challenge directly, acting as a seamless intermediary between enterprise applications and external AI models. It automatically detects and replaces sensitive data, such as names, email addresses, or financial details, with realistic placeholders before any request is sent. When a response returns, the original data is securely restored, ensuring a frictionless user experience without compromising privacy.
“With the rise of AI using closed-source models, employees are pouring sensitive data into AI systems faster than enterprises can secure it,” said Hannes Hapke, Head of Dataiku’s 575 Lab. “That’s not just a technical issue, it’s a governance failure waiting to happen. Kiji Privacy Proxy gives organizations a way to scale AI responsibly, without putting sensitive data and ultimately customer trust on the line.”
Designed for real-world deployment, Kiji Privacy Proxy integrates directly into existing workflows with minimal disruption—whether running as a desktop application for individual users or scaling as a server-based solution for enterprise teams. Unlike traditional approaches, it requires no changes to prompts, applications, or infrastructure, making privacy protection both immediate and practical.
Beyond standard PII detection, Kiji Privacy Proxy allows for domain-level customization. Enterprises can tailor detection to their own data, from proprietary identifiers to highly specific industry- and regional-specific formats, enabling the level of precision and control required in even the most regulated sectors.
Read More: AI Data Infrastructure: The Backbone of AI-Native Enterprises
Kiji Privacy Proxy also signals a broader shift toward open, collaborative AI infrastructure, where trust, transparency, and governance are built in rather than bolted on. Built alongside contributors across the machine learning ecosystem, the project brings together capabilities in workflow orchestration, data annotation, and synthetic data generation to help raise the industry standard for responsible AI. Kiji Privacy Proxy is available now as an open-source project.
Write to us [wasim.a@demandmediaagency.com] to learn more about our exclusive editorial packages and programmes.
