OpenAI releases open-source privacy filter model

OpenAI has launched Privacy Filter, an open-source, on-device model designed to remove personally identifiable information (PII) from enterprise datasets before data is sent to the cloud. This tool, available under a permissive Apache 2.0 license, addresses growing concerns about data privacy and security in AI training and inference. The model, based on OpenAI's gpt-oss family, uses a bidirectional token classification approach and a sparse Mixture-of-Experts architecture for efficiency, allowing it to run on standard hardware with a large context window. OpenAI aims to provide developers with a privacy-by-design toolkit, enabling compliance with regulations like GDPR and HIPAA while still leveraging advanced AI capabilities.