OpenAI Launches Privacy Filter to Protect Sensitive Data Locally
- OpenAI released the Privacy Filter under Apache 2.0, available on GitHub and Hugging Face.
- The model has a size of 1.5 billion parameters and can run locally on regular laptops.
- Privacy Filter achieved a score of 96% F1 on the PII-Masking-300k benchmark.
- It scans for eight categories of personal information, including names, addresses, and passwords.
- The tool is free and open-source, allowing researchers to modify and improve it without legal concerns.
OpenAI’s Privacy Filter is designed to protect sensitive data by masking it before reaching AI tools like ChatGPT. It operates locally, ensuring data privacy by not sending information to cloud services for processing.
With a high accuracy rate of detecting private information at 96%, this tool provides an essential layer of privacy protection for users handling sensitive data daily.(Source)
