OpenAI Expands Zero Data Retention for Frontier Models and Previews Private Safety Processing

What OpenAI announced

OpenAI published an update on 19 August 2026 describing Zero Data Retention support for eligible API customers using frontier models. Under ZDR, OpenAI says prompts and model responses are not retained after request processing, customer content is not available to personnel for routine review, and enterprise customer data is not used for model training unless the customer explicitly opts in.

Private Safety Processing

The company also previewed Private Safety Processing, an approach intended to identify safety-relevant patterns across related interactions without exposing the underlying customer content to OpenAI personnel. OpenAI says the system can operate with content held on customer-controlled infrastructure and is also developing an option where content stored on OpenAI infrastructure is encrypted with keys controlled by the customer.

When automated systems identify a potential risk, the design is intended to return a limited safety signal rather than the underlying prompts or responses. OpenAI says Private Safety Processing is currently being tested with early customers and that it plans to begin rollout and publish a technical white paper in September.

Why it matters

The update addresses a practical tension for enterprises deploying increasingly capable, long-horizon AI systems: safety monitoring may benefit from context across multiple interactions, while regulated or security-sensitive organizations may need strict controls over retention and human access to their data. The announced approach aims to preserve those privacy commitments while extending automated safety monitoring across longer workflows.

Organizations considering ZDR should review OpenAI's official documentation and eligibility requirements before making compliance or architecture decisions.