OpenAI released Privacy Filter, an open-weight model that detects and redacts personally identifiable information in text on-device

OpenAI released Privacy Filter, an open-weight 1.5B parameter model that detects and redacts personally identifiable information across eight categories including names, addresses, phone numbers, emails, account numbers, and passwords, processing text in a single forward pass without sending data to a server. The model achieves a 97.43% F1 score on a corrected version of the PII-Masking-300k benchmark, supports up to 128,000 tokens of context, and is available under the Apache 2.0 license on Hugging Face and GitHub for experimentation, customization, and commercial deployment. OpenAI said it uses a fine-tuned version of Privacy Filter in its own internal privacy workflows and is releasing it to make privacy-preserving infrastructure easier for developers to build into training, logging, and review pipelines.