Under Zero Data Retention, eligible API customers get a contractual guarantee that their prompts and model outputs are discarded once a request completes, are not available to the provider's staff for manual review, and are not used to train future models unless the customer separately opts in. It is common in regulated industries (healthcare, finance, legal) where sending data to a third-party API otherwise creates a retention and access-control liability. OpenAI's August 2026 Private Safety Processing preview is built to keep ZDR intact for frontier models while still catching cross-interaction abuse patterns, using automated pattern detection on customer-controlled or customer-encrypted content instead of retained, human-reviewable logs.