OpenAI expanded Zero Data Retention for frontier models and previewed Private Safety Processing. Here is what it means, and what security teams should verify.
Key takeaways
- On August 19, 2026, OpenAI reaffirmed Zero Data Retention for frontier models: prompts and completions aren't stored after a request or used for training, for eligible API customers.
- OpenAI also previewed Private Safety Processing, a system that checks safety risk across related interactions without personnel accessing your content. It rolls out with a white paper in September.
- This matters because logging and retention were the main blockers stopping regulated teams, in finance, defense, and healthcare, from routing sensitive data through frontier APIs.
- The framing is competitive: per TechCrunch and Axios, OpenAI is contrasting ZDR with Anthropic's retention window for enterprise.
- Van Data Team's recommendation: treat Zero Data Retention as a lever, not a guarantee you can skip verifying. Confirm eligibility, read the terms, and wait for the PSP white paper before relying on it.
📖 Read the full guide on Van Data Team → Zero Data Retention and OpenAI's Private Safety Processing
Top comments (0)