OpenAI has previewed Private Safety Processing (PSP), a system intended to detect safety-relevant patterns across related interactions without giving OpenAI personnel access to the underlying customer content. The company says PSP is designed to remain compatible with its Zero Data Retention (ZDR) commitment.
What ZDR promises
Zero Data Retention (ZDR) gives eligible API customers a clear promise: OpenAI does not retain their prompts or model responses after a request is processed. Customer content is not available to OpenAI personnel for review, and enterprise customer data is not used to train OpenAI models unless customers explicitly opt in.
Why Private Safety Processing is needed
As models take on longer and more complex tasks, some serious risks become apparent only when multiple interactions are considered together. Existing ZDR-compatible safety systems typically evaluate interactions individually. Private Safety Processing is intended to extend those protections across related interactions so that automated systems can identify patterns indicating misuse without exposing retained customer content to OpenAI staff.
Storage options and encryption
PSP can operate regardless of where customer content is stored: either on infrastructure controlled by the customer (ZDR deployments) or on storage provided by OpenAI. When OpenAI-provided storage is used, customer content is encrypted with keys controlled by the customer; OpenAI personnel do not hold copies of those keys and therefore cannot access the underlying content.
Signals, investigations, and customer control
When a risk is detected, OpenAI receives a narrowly defined signal describing the type of activity involved, similar to existing safety signals. That signal can inform whether enforcement action is necessary, but OpenAI personnel do not receive access to the flagged customer content.
Customers can investigate alerts and enforcement decisions using the information available in their own systems. If customers wish to appeal, clarify legitimate activity, or support an investigation into verified abuse, they can choose to share relevant information with OpenAI.
Deployment status and customer involvement
Private Safety Processing is currently being tested with early customers. OpenAI says it announced the preview now because customers have requested predictability about how their content will be protected as AI systems become more capable. The company states it is shaping PSP with feedback from customers across industries, regions, and company sizes.
Legal exception for CSAM
Like other frontier-model providers, OpenAI is legally required to report apparent child sexual abuse material (CSAM). Images flagged for potential CSAM will continue to be retained for manual review and reporting purposes, even in Zero Data Retention deployments.
Customer feedback and rationale
OpenAI emphasizes that collaboration with customers and partners is essential to developing effective safeguards. The organizations working with OpenAI handle highly sensitive information—financial records, health data, confidential business plans, and proprietary research—and protecting that information is crucial for regulatory compliance, customer trust, and competitive advantage.
Sunil Agrawal, Chief Information Security Officer of Glean, commented: “Enterprise AI adoption depends solely on customer control of data, with no direct or derivative use beyond the chosen service. OpenAI’s no-training commitment and ZDR give Glean confidence to build with OpenAI. As models become more capable, OpenAI shows safety can advance without compromising the privacy and control that sustain enterprise trust.”
Next steps and timeline
OpenAI plans to begin rolling out Private Safety Processing and to publish a technical white paper in September. The company says it will keep customers informed throughout the process, explaining implications for existing commitments and providing time and support to plan ahead.
