Source: Data controls in the OpenAI platform
OpenAI's API data-handling guide: no training by default, 30-day abuse-monitoring logs, Modified Abuse Monitoring and Zero Data Retention, endpoint eligibility, application-state retention and EKM.
Contents
- What: official API documentation page
- Publisher: OpenAI
- Date: accessed 2026-11-14
- Original: Data controls in the OpenAI platform
Summary
Describes what OpenAI keeps when you call the API, for how long, and which controls reduce it. Two separate stores matter: abuse-monitoring logs, and application state that some endpoints keep so that they can work (stored responses, files, vector stores).
Key claims
- API data has not been used for training since 2023-03-01 unless the customer opts in.
- Abuse-monitoring logs are kept for up to 30 days by default, longer when the law requires it.
- Modified Abuse Monitoring (MAM) removes customer content from those logs. Zero Data Retention (ZDR) does the same and also forces
store=falseon/v1/responsesand/v1/chat/completions. Both need prior approval and are set per organization, with per-project override (a project can also choose None). - ZDR-eligible endpoints include responses, chat completions, embeddings, audio, moderations, completions and realtime. Not eligible: conversations, ChatKit threads, agents, assistants, threads, vector stores, audio voices, files, fine-tuning, evals, batches.
- Responses with
storeomitted or true are kept at least 30 days. Files are kept until deleted. Deleted assistants, threads and vector stores are purged 30 days later. Prompt-cache tensors can stay in GPU-local storage up to 24 hours. - Flagged content can be kept for review even under ZDR or MAM, and image and file inputs are scanned for CSAM.
- Enterprise Key Management encrypts customer content with keys held in AWS KMS, Google Cloud or Azure Key Vault.