A local-first, reversible PII scrubber for AI workflows
Hi HN, I’m one of the maintainers of Bridge Anonymization. We built this because the existing solutions for translating sensitive user content are insufficient for many of our privacy-concious clients (Governments, Banks, Healthcare, etc.). We couldn't send PII to third-party APIs, but standard redaction destroyed the translation quality. If you scrub "John" to "[PERSON]", the translation engine loses gender context (often defaulting to masculine), which breaks grammatical agreement in languages like French or German. So we built a reversible, local-first pipeline for Node.js/Bun. Here…
In plain words
Bridge Anonymization is a local-first tool for Node.js and Bun that removes personally identifiable information from text before processing by AI translation systems. Unlike standard redaction methods that replace names with generic tokens, it preserves linguistic context needed for accurate translation in gendered languages. The tool keeps PII local, never sending sensitive data to third-party APIs, making it suitable for government, banking, and healthcare organizations that require strict privacy controls.
written from the facts on this page · September 2026
From the sources
In the maker’s words, at launch
Hi HN, I’m one of the maintainers of Bridge Anonymization. We built this because the existing solutions for translating sensitive user content are insufficient for many of our privacy-concious clients (Governments, Banks, Healthcare, etc.). We couldn't send PII to third-party APIs, but standard redaction destroyed the translation quality. If you scrub "John" to "[PERSON]", the translation engine loses gender context (often defaulting to masculine), which breaks grammatical agreement in languages like French or German. So we built a reversible, local-first pipeline for Node.js/Bun. Here is how we implemented the tricky parts: 0. The Mapping We use XML-like tags with ID’s that uniquely identify the PII, `<PII type=”PERSON” id=”1”>`. Translation models and the systems around them work with XML data structures since the dawn of Computer Aided Translation tools, so this improves compatibility with existing workflows and systems. A `PIIMap` is stored locally for rehydration after translation (AES-256-GCM-encrypted by default). 1. Hybrid Detection Engine Obviously neither Regex nor NER was enough on its own. - Structured PII: We use strict Regex with validation checksums for things like IBANs (Mod-97) and Credit Cards (Luhn). - Soft PII: For names and locations, we run a quantized `xlm-roberta` model via `onnxruntime-node` directly in the process. This lets us avoid a Python sidecar while keeping the package ‘lightweight’ (still ~280MB for the quantized model, but acceptable for desktop environments). 2. The "Hallucination" Guard (Fuzzy Rehydration) LLMs often "mangle" the XML placeholders during translation (e.g., turning `<PII id="1"/>` into `< PII id = « 1 » >`). We implemented a Fuzzy Tag Matcher that uses flexible regex patterns to detect these artefacts. It identifies the tag even if attributes are reordered or quotes are changed, ensuring we can always map the token back to the original encrypted value. 3. Semantic Masking We are currently working on "Semantic Masking"—adding context to the PII tag (like `<PII type="PERSON" gender="female" id="1" />` ) to preserve (gender) context for the translation. For now, we are relying on a lightweight lookup-table approach to avoid the overhead of a second ML model or the hassle of fine tuning. So far this works nicely for most use cases. The code is MIT licensed. I’d love to hear how others are handling the "context loss" problem in privacy-preserving NLP pipelines! I think this could quite easily be generalized to other LLM applications as well.
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 26d ago · cactuscompute.com
Launched alongside, December 2025
the whole month →- GP
Life & fun · Dec 2025 · dosaygo-studio.github.io
- JG
Hi everyone! My name's Luke and I made the original Jmail here alongside Riley Walz. We had a ton of friends collaborate on building out more of the app suite last night in lieue of DOJ's "Epstein files" release. Please AMA!
Life & fun · Dec 2025 · jmail.world
PlanEat AI▲733AI turns your health goals into a 7-day menu & grocery list
AI · Dec 2025 · planeatai.com
- 2G
Community, All the HN belong to you. This is an archive of hacker news that fits in your browser. When I made HN Made of Primes I realized I could probably do this offline sqlite/wasm thing with the whole GBs of archive. The whole dataset. So I tried it, and this is it. Have Hacker News on your device. Go to this repo (https://github.com/DOSAYGO-STUDIO/HackerBook): you can download it. Big Query -> ETL -> npx serve docs - that's it. 20 years of HN arguments and beauty, can be yours forever. So they'll never die. Ever. It's the unkillable static archive of HN and it's…
Dev tools · Dec 2025 · hackerbook.dosaygo.com
