The Local Guardian: OpenAI Unveils a 1.5B Open-Source Model to Redact PII Locally

OpenAI's Privacy Filter is an open-source 1.5B-parameter autoregressive model engineered to detect and tag personally identifiable information (PII) in textual inputs for local redaction; it operates as a tagging mechanism (requiring additional rules or models to remove or replace flagged spans), supports up to a 128K token context window, and is optimized for execution on consumer-grade hardware such as web browsers.