OpenAI Expands Zero Data Retention to Keep Enterprise Customer Data Private
Eligible API customers can now run requests through frontier models without OpenAI saving their inputs or outputs.
OpenAI just expanded Zero Data Retention policies to cover its most capable frontier models. In plain English, that means if you use their API — the software pipeline that lets your apps talk directly to OpenAI's models — your prompts and generated answers are not saved to disk or used for training future systems.
The company also previewed what it calls Private Safety Processing. Previously, even when companies opted out of data storage, AI providers had to run automated abuse-monitoring checks that temporarily held onto data. This new setup allows security and safety filters to run in isolated environments without compromising customer confidentiality.
The Race for Enterprise Trust
This move turns up the heat on Anthropic, which has historically won over security-conscious enterprise teams by emphasizing strict data governance. When big businesses handle sensitive legal contracts, proprietary code, or personal user records, privacy is non-negotiable.
What This Means for You as a Builder
If you are pitching custom AI agents to clients in healthcare, legal, or finance, data privacy is almost always their very first objection. Knowing that you can offer Zero Data Retention out of the box makes it far easier to get internal stakeholders on board without getting stuck in months of security reviews.