OpenAI Zero Data Retention: How to Keep Your API Data Private

What is OpenAI Zero Data Retention and what changed?

OpenAI Zero Data Retention (ZDR) is a privacy commitment ensuring that prompts and model responses aren’t stored after a request is processed.

Eligible API customers receive a guarantee that their content isn’t available to OpenAI personnel for review. Enterprise customer data isn’t used to train models unless the customer explicitly opts in.

The August 19, 2026 update previews Private Safety Processing, a system that identifies misuse patterns across multiple interactions without granting OpenAI staff access to the underlying data.

ZDR plus Private Safety Processing gives eligible API customers a clear promise that their content stays under their control.

What is the evidence behind OpenAI Private Safety Processing?

Private Safety Processing uses automated systems to detect harmful intentions that only become visible across multiple interactions, such as bad actors probing safeguards or AI agents becoming misaligned.

The system works whether content stays on customer-controlled infrastructure or is stored by OpenAI. In the OpenAI-stored case, data is encrypted with keys controlled by the customer, and OpenAI personnel don’t have copies of those keys.

When a risk is flagged, OpenAI receives a narrowly defined signal indicating the activity type. Personnel still don’t receive access to customer content even when a flag is triggered.

Automated safety signals replace the need for human review of sensitive customer prompts.

How does OpenAI Zero Data Retention compare to the alternatives, and what background do small business owners need?

Other frontier model providers require customers to allow data retention for safety monitoring, which conflicts with the security obligations of organizations handling financial records or health data.

ZDR removes this conflict by keeping data on infrastructure the customer controls, or by storing it on OpenAI infrastructure with customer-held encryption keys.

Customers can investigate alerts and enforcement decisions using information available in their own systems. They can choose to share relevant information with OpenAI only for appeals, clarification, or investigations into verified abuse.

ZDR keeps regulatory conflicts out of the AI procurement conversation for organizations handling sensitive records.

The intake form sits open on the front desk of the medical clinic, and a patient just handed over their full health history because the new specialist needs it before the appointment. You scan the documents, paste the file into your AI assistant to draft a summary, and the API provider logs every prompt you just sent.

The compliance officer calls 2 hours later asking why sensitive records are sitting in a third-party training pipeline. With ZDR eligible on your API plan, that call never happens, because OpenAI doesn’t retain the prompts after processing and personnel can’t open them.

Private Safety Processing rolls out in September, which means even the misuse-detection layer no longer requires humans to read your data. That’s the structural shift for any founder who handles other people’s sensitive information.

How does OpenAI Zero Data Retention affect day-to-day operations for small businesses?

Small businesses in regulated industries can integrate frontier models into internal workflows without violating compliance obligations. Law firms can process confidential case files and health clinics can analyze patient data through the API, which connects to broader operational intelligence signals tracked across the Wire.

The only exception is child sexual abuse material (CSAM), which OpenAI continues to retain for manual review and reporting purposes, even in ZDR deployments, as required by law.

Financial advisors can run proprietary research through the API without leaking client strategies into a shared model. Health clinics can analyze patient data with the same protections they apply to HIPAA-covered records.

Legal practices can process confidential case files and attorney work product through the API. None of these workflows require the AI provider to retain the prompts or have personnel review them.

Private Safety Processing is currently being tested with early customers. OpenAI plans to start rolling it out, and share a technical white paper, in September. The white paper will detail how the safety signals work without exposing customer content to OpenAI personnel.

Day-to-day operations gain a defensible privacy posture without losing access to frontier models.

What is the final verdict on OpenAI Zero Data Retention?

The September rollout of Private Safety Processing ensures that safety monitoring doesn’t come at the cost of customer privacy. It allows deployment of complex AI agents that can operate within the bounds of their intended authority.

Founders handling financial records, health data, confidential business plans, or proprietary research should audit their current API settings to confirm ZDR eligibility before scaling agentic workloads.

OpenAI named Glean, Databricks, Abridge, and Microsoft as early collaborators, which signals that enterprise-grade trust is being built into the deployment model.

ZDR with Private Safety Processing is the privacy baseline for any small business running sensitive data through frontier AI.

Source: OpenAI Blog