Piiranha v1 is currently the best privacy AI model available on HuggingFace. It was trained on H100s from the Akash Supercloud by @itsandrewgao. https://t.co/lnadvdCKlG
We trained the best Privacy AI model available on @huggingface: Piiranha fully open source and thoroughly documented, so you can actually use it. recognizes 17 types of personal data trained on H100s from @akashnet_ https://t.co/6tdKYPId3J
We noticed the lack of a performant, *documented* Personally Identifiable Information model on @huggingface So, we fine-tuned Deberta-V3-base on nearly 2 million PII examples. Unlike other PII detectors on Hugging Face, Piiranha is actually usable because we: report… https://t.co/seqC8bhcwC https://t.co/6JdZsxW8Wk
Arcjet has introduced a new sensitive information detection tool that detects, blocks, and redacts Personally Identifiable Information (PII) submitted to applications and LangChainAI chat interfaces. This tool operates locally, ensuring that no data leaves the user's environment. In a related development, a new PII detection model named Piiranha has been fine-tuned by Andrew Gao and is available on Hugging Face. Piiranha is open-source, thoroughly documented, and trained on nearly 1.8 million PII examples, recognizing 17 types of personal data. It boasts a 98.27% detection rate with only 0.16% false positives. The model was trained on H100s from the Akash Supercloud and is based on Deberta-V3-base. Users can access the model's colab and repo for implementation.