OpenAI Releases Privacy Filter: A 1.5B-Parameter Open-Source PII Redaction Model with 50M Active Parameters

OpenAI released Privacy Filter, an Apache 2.0 open-source model for detecting and redacting PII.

It can identify eight categories of sensitive data using sparse MoE and bidirectional token classification.

The model has 1.5B total parameters but only 50M active at inference, helping it run efficiently on local or edge hardware.

By keeping redaction on-device, teams can scrub logs, datasets, and user content without sending data to a third-party API.