OpenAI previews Private Safety Processing to support Zero Data Retention for frontier models

OpenAI previewed Private Safety Processing, a system designed to detect patterns of potential misuse across related interactions while remaining compatible with Zero Data Retention (ZDR). Customer content stays under customer control or is encrypted with customer-held keys so OpenAI personnel cannot access the underlying prompts or responses. The approach aims to strengthen multi-turn and agentic safety monitoring without requiring content retention that conflicts with enterprise security and privacy obligations. Rollout and a technical white paper are planned for September.