OpenAI is expanding its Zero Data Retention (ZDR) offering for eligible API customers, now previewing Private Safety Processing to detect misuse patterns across related interactions without exposing customer content to OpenAI personnel.

For ZDR deployments, customer content remains on infrastructure the customer controls. OpenAI is also developing an encrypted storage option where content is stored on OpenAI infrastructure but encrypted with customer-controlled keys.

Private Safety Processing builds on existing ZDR-compatible safety systems by evaluating interactions collectively rather than individually. When risks are identified, OpenAI receives narrowly defined signals about activity types without accessing underlying prompts or responses.

The system is currently being tested with early customers and will roll out in September, along with a technical white paper. OpenAI emphasizes collaboration with enterprise customers handling sensitive data across financial, health, and research sectors.