OpenAI's New Privacy Filter Model Released

OpenAI has released Privacy Filter, an open-source model designed to redact PII from text. The model is lightweight with a sparse design, allowing on-premises deployment without external data sharing.

Published Apr 29, 2026, 3:58 AMUpdated Apr 29, 2026, 3:58 AM

What happened

OpenAI released Privacy Filter, an open-source PII redaction model with 1.5 billion parameters, but only 50 million active parameters, making it suitable for on-premises deployment.

[1]

Why it matters

The model facilitates secure data handling by enabling on-premises PII redaction without requiring data to be sent to third-party APIs, helping organizations comply with privacy regulations.

[1]

Who is affected

Developers and organizations handling sensitive data can use this model to sanitize datasets while maintaining data privacy, especially in regulated industries.

[1]

Risks / uncertainty

The model has known failure modes, like missing novel credential formats, which may affect its reliability in certain contexts.

[1]