Edward Coristine and Tai Groot built the Hugging Face-hosted model.
Today, we are releasing Rampart: a 14.7MB machine learning model designed to protect citizens’ privacy by redacting personal information directly in your browser before it gets sent to any server
A big release of something small is coming tomorrow. Live at 12:30PM EST. https://x.com/i/broadcasts/1dKrPPRmgQXJX
Many users praised NDStudio's Rampart open-source model for enabling local in-browser privacy redaction without sending data elsewhere, while some criticized the team or called the focus insufficient.
No Digg Deeper questions have been answered for this story yet.
Tired: The US government regulating open-source AI models Wired: The US government training and releasing open-source AI models
Today, we are releasing Rampart: a 14.7MB machine learning model designed to protect citizens’ privacy by redacting personal information directly in your browser before it gets sent to any server
Introducing Rampart, an open source model to protect the privacy of citizens’ personal information, built by the talented software team at @ndstudio. Watch the recorded live stream below with @as400495 @taigrr 🇺🇸
Today, we are releasing Rampart: a 14.7MB machine learning model designed to protect citizens’ privacy by redacting personal information directly in your browser before it gets sent to any server
You wake up. It’s 2026. The newest open source AI model was trained by federal government employees, Big Balls and Groot.
FULL INTERVIEW: Engineers Edward Coristine (@as400495) and Tai Groot (@taigrr) just released an ML model called Rampart for the National Design Studio.
It's a local-first, open-source AI privacy model that runs entirely in the browser.
They joined TBPN to explain the model and why the National Design Studio launched it.
0:25 – National Design Studio launches Rampart 1:40 – Why Rampart is open source 2:40 – How NDS trained the model 3:55 – Use cases for the model 5:25 – How NDS thinks about side quests 6:25 – Team size, workflow, and why they work seven days a week 8:00 – Recruiting top talent to the National Design Studio
🎯
FULL INTERVIEW: Engineers Edward Coristine (@as400495) and Tai Groot (@taigrr) just released an ML model called Rampart for the National Design Studio.
It's a local-first, open-source AI privacy model that runs entirely in the browser.
They joined TBPN to explain the model and why the National Design Studio launched it.
0:25 – National Design Studio launches Rampart 1:40 – Why Rampart is open source 2:40 – How NDS trained the model 3:55 – Use cases for the model 5:25 – How NDS thinks about side quests 6:25 – Team size, workflow, and why they work seven days a week 8:00 – Recruiting top talent to the National Design Studio
Very cool work @jgebbia and team.
Today, we are releasing Rampart: a 14.7MB machine learning model designed to protect citizens’ privacy by redacting personal information directly in your browser before it gets sent to any server
@ndstudio I wouldn't trust these people. https://thedreydossier.github.io/NDS_servers_map/
HuggingFace: https://huggingface.co/nationaldesignstudio/rampart
FULL INTERVIEW: Engineers Edward Coristine (@as400495) and Tai Groot (@taigrr) just released an ML model called Rampart for the National Design Studio.
It's a local-first, open-source AI privacy model that runs entirely in the browser.
They joined TBPN to explain the model and why the National Design Studio launched it.
0:25 – National Design Studio launches Rampart 1:40 – Why Rampart is open source 2:40 – How NDS trained the model 3:55 – Use cases for the model 5:25 – How NDS thinks about side quests 6:25 – Team size, workflow, and why they work seven days a week 8:00 – Recruiting top talent to the National Design Studio
Tomorrow, we're releasing an ML model for on-device PII redaction, which is far smaller than existing models, but is surprisingly effective. It is a first iteration. Will be live at 12:30!
@ndstudio How long until we hear about the back door tool calling they bake in.
@ndstudio how does it stack up against GLiNER or Presidio?
Read more on our dev index: https://ndstudio.gov/posts/say-hello-to-rampart
The raw text you enter into a chatbot might include names, addresses, or financial details. Rampart detects and replaces sensitive terms with stable placeholders like [NAME_1], and sends the cleaned version instead. This way, the model still understands your request but your data doesn't have to leave the device
@ndstudio what would be super helpful would be a version that could anonymise large legal docs (e.g. contracts, shareholder resolutions etc.) with stable placeholders that span across docs - anything like this planned?
This is our first iteration of Rampart. We built it because we wanted Americans to have better PII removal on-device along with other privacy preserving measures. If you are motivated by the work of building better technology to serve Americans, consider joining National Design Studio.
@ndstudio Ok. So you rebuilt PII Shield that’s been open source for more than a month. Well done.
HuggingFace: https://huggingface.co/nationaldesignstudio/rampart
@ndstudio Awesome
Rampart is trained on synthetic data, open source datasets, and government FOIA releases, including OCR-processed documents to improve performance on noisy data
@ndstudio Love that this has a TS lib, link for those curious: https://github.com/nationaldesignstudio/rampart
Read more on our dev index https://ndstudio.gov/posts/say-hello-to-rampart
