OpenAI unveils Privacy Filter to erase PII on-device before any cloud upload

A 128k context model that redacts names precisely

OpenAI has released Privacy Filter, an open-source model that detects and redacts personally identifiable information before data ever leaves an enterprise environment. Built from a gpt-oss variant, it runs locally on laptops or in browsers and supports large 128,000-token inputs. The tool targets fast, high-throughput privacy pipelines, but comes with a caution against treating it as a full safety guarantee.