OpenAI Private Intelligence Brings New Data Controls to Frontier AI Deployments

OpenAI's Private Intelligence initiative aims to provide businesses with a more defined set of controls for data privacy when deploying its frontier AI models. This offering is presented as a formal product, rather than just an updated privacy statement, intended to help companies manage sensitive operational information. It seeks to separate safety reviews from direct access to underlying content, while also outlining future capabilities for confidential inference.

The initiative includes several components with differing availability. Private Safety Processing (PSP) is described as immediately available for relevant API customers under Zero Data Retention terms. OpenAI states that PSP enables automated safety reviews without allowing its personnel to access the underlying content. This design point is particularly relevant for API-based use cases involving business material, providing a distinct approach to processing and human access.

However, other key features remain on OpenAI's development roadmap. Private Inference, which OpenAI describes as a confidential-computing capability with strict controls, is scheduled for a preview in fall 2026, indicating it is not yet generally available. Similarly, an optional Enterprise Key Management (EKM) layer is part of the broader package, but the announcement does not specify its operational details, eligible customers, or pricing. Consequently, organizations evaluating Private Intelligence for deployment must differentiate between currently available controls and those still in development or early access.