Hugging Face CEO Clément Delangue sees the White House-linked Rampart model claiming the top trending spot in token classification as proof that governments prefer to control their own model weights rather than depend on commercial providers.
Rampart from the @ndstudio @WhiteHouse is number one trending token classification model on HF.
Very cool to see public organizations starting to own and build their weights instead of renting them from an API provider!
https://huggingface.co/models?pipeline_tag=token-classification&sort=trending
The 14.7 MB quantized ONNX model detects 17 PII entity types across seven Latin-script languages and redacts them in the browser before data reaches servers, released under CC BY 4.0 with an accompanying npm package.
Created by the National Design Studio formed via 2025 executive order, the model is presented as one concrete case of a government body training and sharing its own weights on an open platform.
Many users praised the White House for building its own Rampart model weights on Hugging Face instead of renting from OpenAI, calling the independence a huge and necessary shift.
No Digg Deeper questions have been answered for this story yet.
🇺🇸🇺🇸 @ndstudio
Rampart from the @ndstudio @WhiteHouse is number one trending token classification model on HF.
Very cool to see public organizations starting to own and build their weights instead of renting them from an API provider!
https://huggingface.co/models?pipeline_tag=token-classification&sort=trending
@ClementDelangue @ndstudio @WhiteHouse the white house owning its own weights before half the fortune 500 is crazy work ngl
@ClementDelangue @ndstudio @WhiteHouse nature is healing. the white house is on huggingface instead of paying openai bills.
@ClementDelangue @ndstudio @WhiteHouse honestly love this.
@ClementDelangue @ndstudio @WhiteHouse Renting weights is such a lazy pattern. Owning them means you actually understand what youre building
@ClementDelangue @ndstudio @WhiteHouse public orgs building weights instead of renting is huge
api dependency is a silent tax on innovation and control
@ClementDelangue @ndstudio @WhiteHouse ownership matters most when the dependency starts looking like infrastructure.
@ClementDelangue @ndstudio @WhiteHouse Unfortunate that it’s English only, we decided to make one that works great with all 24 EU languages
https://huggingface.co/spaces/desert-ant-labs/redact-demo
@jgebbia @ndstudio Can’t wait to get in the office and start using it!
@ClementDelangue @ndstudio @WhiteHouse building their own weights instead of renting is exactly the kind of shift we need
curious what the inference cost difference looks like vs an api
@jgebbia @ndstudio Great work
@jgebbia @ndstudio Love this
@jgebbia @ndstudio Gallantly streaming. 🇺🇸