GoKawiilNational Design Studio open-sources Rampart, a browser-based PII redaction tool
National Design Studio released Rampart, an open-source system that detects and strips personal information from text before it leaves a user's device. It runs entirely in the browser, combining regex-based rules for structured data like Social Security numbers with the MiniLM machine learning model for names and addresses, aiming to complete the process in milliseconds.
GoKawiil's interpretation of the reporting above, not reported fact.
By processing data locally rather than on a remote server, Rampart sidesteps the need to trust AI vendors' privacy claims, which the developers argue are inherently unverifiable once data leaves a user's device. Its lightweight design also contrasts with larger server-side filters like OpenAI's, which the company says can be too large to practically deploy to average users' browsers, potentially making privacy protection more accessible on constrained connections.
- Rampart performs personal information redaction entirely on-device, inside the browser.
- It uses a hybrid approach: deterministic regex rules plus the MiniLM model for names and addresses.
- The project is open source, positioned as a lightweight alternative to large server-based PII filters.
Source: ndstudio.gov, 2026-10-09
Published there as: “Rampart: Browser native on-device PII radaction”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.