OpenAI Releases Open Source Model OpenAI Privacy Filter, Capable of Detecting and Desensitizing Personal Privacy Information in Text

Odaily News OpenAI today released the open-source model OpenAI Privacy Filter, designed to detect and redact personally identifiable information (PII) in text. The model boasts a total of 1.5 billion parameters with 50 million active parameters, supporting a context window of up to 128,000 tokens. OpenAI Privacy Filter employs a bidirectional token classification model architecture, capable of identifying eight categories of information including private names, addresses, emails, phone numbers, URLs, dates, account numbers, and keys, achieving a 96% F1 score on the PII-Masking-300k benchmark. The model is now available on Hugging Face and GitHub under the Apache 2.0 license, supporting local deployment and fine-tuning by developers.