OpenAI has reiterated its Zero Data Retention policy for qualifying API users, ensuring that prompts and outputs are not stored after processing. This applies to customers meeting certain eligibility criteria, reinforcing the company's commitment to data‑privacy guarantees in its API offering.
In parallel, OpenAI is previewing a feature called Private Safety Processing, which runs safety checks on model outputs within a secure enclave so that raw data never leaves the user's environment. The approach aims to maintain high safety standards while preserving confidentiality, representing an architectural shift toward on‑premise‑style safety evaluation.
- Zero Data Retention applies to eligible API customers, preventing storage of prompts and completions.
- Private Safety Processing runs safety checks inside a trusted execution environment, keeping data local.
- The preview signals a move toward privacy‑preserving safety mechanisms without sacrificing model utility.
Why this matters
Source fact: OpenAI has confirmed that eligible API customers receive Zero Data Retention and is testing Private Safety Processing to keep data within user‑controlled environments. Inference: This move likely addresses growing regulatory pressure and enterprise concerns about data exposure, potentially making OpenAI's models more attractive for sectors with strict compliance requirements.