tech
Offering Zero Data Retention for frontier models
Previewing Private Safety Processing, which strengthens safeguards across interactions while remaining compatible with ZDR.
TL;DR
- OpenAI is previewing Private Safety Processing, a new system to strengthen AI safeguards.
- Private Safety Processing is designed to work with Zero Data Retention (ZDR), ensuring OpenAI does not retain customer prompts or model responses.
- The system identifies patterns across related interactions to detect potential misuse that might not be visible in single interactions.
- Customer content can remain on customer-controlled infrastructure or be stored on OpenAI infrastructure, encrypted with customer-controlled keys.
- Automated systems detect risks and return safety signals without exposing raw content to OpenAI personnel.
- Customers can investigate alerts and choose to share information with OpenAI for appeals or investigations.
- This development aims to meet customer needs for data privacy and control as AI models become more capable.