OpenAI Introduces Zero Data Retention Service for Frontier Models
Maximizing customer privacy, OpenAI has begun testing new security processes that eliminate data retention requirements.
OpenAI has announced a Zero Data Retention policy and a Private Safety Processing application where prompts or model responses are not stored for eligible API customers.
Zero Data Retention Policy
Zero Data Retention provides eligible API customers with clear assurance that prompts or model responses are not stored after a request is processed.
Privacy and Data Security
Customer content is not open to review by OpenAI personnel, and enterprise customer data is not used to train models unless explicit consent is given.
Private Safety Processing Procedure
OpenAI has previewed the Private Safety Processing application, designed to identify security patterns between relevant interactions without allowing employees access to the underlying content.
Infrastructure and Encryption Options
In Zero Data Retention deployments, customer content remains directly on the infrastructure controlled by the customer. Furthermore, it was stated that the company is working on an option where content will be stored on OpenAI infrastructure encrypted with customer-controlled keys.
Testing Process and Future Plans
Private Safety Processing is currently being tested with early-stage customers, with a public rollout and a technical white paper planned for September.