OpenAI Launches Private Intelligence, a Set of Data Protections for Businesses Using Its Top Models
Privacy should advance as fast as intelligence. OpenAI Private Intelligence gives businesses stronger controls over their data while using our most capable models. We’re building these protections in close collaboration with some of our largest customers to help businesses apply frontier AI to their most sensitive work.
Context
Stronger guarantees about data access could unlock frontier AI for banks, hospitals and other firms that have held back over confidentiality, a key battleground with Anthropic for enterprise customers.
Zero Data Retention is an OpenAI option where prompts and responses are not kept after a request is processed. Confidential computing keeps data encrypted even while it is being processed.
Deep dive
OpenAI said on Friday that privacy should advance as fast as intelligence, introducing OpenAI Private Intelligence. According to the post, it gives businesses stronger controls over their data while using OpenAI's most capable models, and is being built in close collaboration with some of its largest customers.
OpenAI's product page groups several protections under the name. It says OpenAI sets strict limits on its own access to business content and offers Zero Data Retention and Private Inference for eligible workflows. Businesses own their inputs and outputs, and they decide what agents can access, which actions they can take and when approval is required. Data is protected with encryption, access controls and threat detection.
OpenAI's interest form says Zero Data Retention with Private Safety Processing allows automated safety reviews without OpenAI staff seeing the underlying content. Private Inference, available as an API preview, combines confidential computing with strict, verifiable controls.
OpenAI did not publish pricing or a general availability date, and businesses must register interest. The launch targets regulated industries where data handling rules have slowed adoption of frontier models, as its models take on longer agent tasks.
What to expect next
Watch for when Private Inference leaves preview and for technical documentation showing how its controls can be verified. Large customers in finance, health and government will be the main test. Anthropic, Google and Microsoft are likely to answer with their own confidential computing offers as privacy becomes a selling point for enterprise AI. Customers will also want independent audits showing that staff truly cannot see their content.