Image: OpenAI

OpenAI is promoting a set of business privacy protections under the name OpenAI Private Intelligence. In a post on X, the company said the program gives businesses stronger controls over their data while they use its most capable models, and that it is building the protections in close collaboration with some of its largest customers.

The Private Intelligence page pitches the effort under the line "Privacy should advance as fast as intelligence". OpenAI groups the protections into four areas. On access, it says it sets strict limits on its own access to business content. On ownership, it says businesses own their data, including inputs and outputs. On agents, it says customers decide what AI agents can access, which actions they can take and when they need approval. On security, it cites encryption, strict access controls and ongoing threat detection.

Two technical pieces sit at the center. The first is Zero Data Retention, which OpenAI says is available for eligible API workflows and means it does not keep prompts or model responses after processing. The second is Private Inference, which the page labels as a preview. OpenAI describes it as combining confidential computing with strict controls designed to protect prompts and responses while the model processes them. Confidential computing generally refers to running workloads inside hardware protected environments that even the cloud operator cannot easily inspect.

An OpenAI interest form adds one more detail. It says Zero Data Retention with Private Safety Processing allows automated safety reviews without giving OpenAI personnel access to the underlying content. That addresses a long standing tension for enterprise customers, since providers typically keep some data to check for abuse.

The page also lists existing compliance credentials, including SOC 2 Type 2 audits, ISO certifications, HIPAA support through Business Associate Agreements for eligible services and FedRAMP support for designated government offerings. It features customer videos from Glean and Coinbase.

The program is not entirely new. VentureBeat reported that OpenAI first unveiled Private Intelligence at its DevDay 2026 event, with Private Inference described at the time as coming this fall. According to that report, content flagged for safety review under Private Safety Processing is encrypted and stored in storage controlled by the customer, such as AWS S3, Azure Blob or Google Cloud Storage, rather than kept by OpenAI in plain text.

Outside observers have pointed to open questions. VentureBeat noted that OpenAI had not said which models and API endpoints Private Inference will support, what hardware it will use, what customers will be able to verify independently or how much it will cost. Security engineer Pawan Khandavilli, in a detailed analysis of OpenAI's published white paper, wrote that the safety processing system is the more mature piece while Private Inference remains a bet whose trust boundary has not yet been published.

At the time of this writing, OpenAI had not announced pricing or a general availability date for Private Inference. The company is directing interested businesses to its interest form and Trust Portal.

Privacy should advance as fast as intelligence. OpenAI Private Intelligence gives businesses stronger controls over their data while using our most capable models. pic.twitter.com/RWhIX72c09

— OpenAI (@OpenAI) October 30, 2026