OpenAI Releases Privacy Filter Model to Redact Sensitive Data - Bloomberg Law News

OpenAI Inc. released a customizable AI model on Wednesday called the OpenAI Privacy Filter. The model is designed to identify and redact personally identifiable information such as names and bank account numbers from text. According to the company, this includes text from AI models' training data.

The model is open-weight, meaning its architecture and weights are publicly available. OpenAI stated the release is part of its broader effort to provide developers with tools for building AI safely. The announcement did not specify the model's performance metrics, training data, or immediate availability for integration.

The release comes amid increasing regulatory and public scrutiny over data privacy in AI development. Tools for redacting sensitive information are seen as one technical approach to addressing concerns about data used to train large language models.