Local encoders that strip 40 kinds of PII before text reaches a model
- Source
- x.com
- Date
Sharing confidential information with AIs is a norm these days! We can do better. Our latest encoder models find and strip 40 types of PII across 16 languages, all in one simultaneous pass for a secure information sharing with AIs. LFM2.5-Encoders identify all entities in the text at once. You can choose the entities yourself. They run inside your pipeline, so your private data never leaves your environment. No generation, nothing to parse.
Try them out: LFM2.5-Encoder-230M: https://t.co/RlqT4fFh9Y LFM2.5-Encoder-350M: https://t.co/PR5WQuTY0L
Context
Liquid AI describes a personal-information detector that can sit before an AI service in a text-processing pipeline. Its fine-tuned encoders identify selected categories of sensitive information together, allowing detected passages to be removed without generating a written response. The company advertises coverage of 40 categories across 16 languages. Announcement.
The model card identifies this as a fine-tuned application of the encoder. Running that detector within an organization's own environment can keep the original text there while redacted text goes onward. Local processing controls where detection happens; the advertised category count does not establish that every sensitive passage will be caught.
postLiquid AI ships LFM2.5-24B-A2B open weights for local agents
postToday we release Open d1: two open-weight multimodal models in our d1 decision…
postLiquid AI's d1 decision model adds vision at a fraction of frontier cost
articled1 is now available on @vercel AI gateway!We worked with the amazing vercel…
Checking sign-in…
Loading comments…