02
OpenAI keeps zero data retention for frontier models, splitting with Anthropic
verifiedLegalEnterprise
Thursday, August 20, 2026
Confidence
High · — OpenAI primary + 4 corroborating; September white paper still pending
Evidence
OpenAI newsroom post + Axios reporting + rival policy on record
OpenAI committed that eligible API customers keep zero data retention on frontier models — no retained prompts, no personnel access, no training use without opt-in — and previewed Private Safety Processing.
- Detects misuse patterns across related interactions without exposing content to staff.
- Microsoft and Databricks are early customers; white paper due September.
Sources