OpenAI reaffirmed Zero Data Retention for eligible API customers and previewed Private Safety Processing for advanced frontier models

Decision brief

What changed

OpenAI said eligible API customers can continue to use Zero Data Retention services, where customer data is not retained for those services, and it previewed Private Safety Processing for advanced frontier models as a customer-preview option to monitor misuse without retaining customer data.

Why it matters

This matters for leaders deciding whether sensitive workloads can be moved onto frontier-model APIs without changing internal data-handling requirements. It also suggests OpenAI is trying to preserve privacy commitments while adding stronger misuse monitoring on advanced models, which affects vendor selection, compliance review, and deployment design for high-risk use cases.

Affected roles

CEO CTO CISO COO

Evidence

Two OpenAI blog posts state that OpenAI reaffirmed Zero Data Retention for eligible API customers and previewed Private Safety Processing as a privacy-focused safety option for frontier models. TechCrunch independently reported the same core announcement and added context that OpenAI positions this as an expansion of ZDR, with misuse monitoring and no customer-data retention.

What remains uncertain

The coverage does not specify eligibility criteria, product availability, pricing, technical limitations, or which frontier models are included, so operational impact cannot yet be fully assessed. Claims about broader long-horizon monitoring and how enforcement works come from media characterization and remain dependent on OpenAI’s eventual technical and policy details.

Monitor next

Watch for OpenAI to publish formal eligibility rules, model coverage, and generally available documentation for Private Safety Processing, including enforcement workflows and any compliance attestations.

Analytical support, not advice — assumptions and open questions stated above.

Source coverage