OpenAI Launches Private Intelligence to Win Over Data-Wary Enterprise Customers

Quick Facts

OpenAI unveiled Private Intelligence at DevDay 2026, an umbrella initiative combining Zero Data Retention with Private Safety Processing, available now for API customers.
A second technology, Private Inference, will run model inference inside hardware-sealed enclaves and is set for a fall 2026 preview.
The move directly counters Anthropic, which began requiring mandatory 30-day data retention for business clients using its most powerful models in June 2026.

OpenAI is building a direct answer to one of enterprise AI's most persistent problems: who can actually see your data. At DevDay 2026, the company announced Private Intelligence, an umbrella program that bundles existing and new privacy-preserving tools for business and API customers.

The initiative has two components. The first, Zero Data Retention with Private Safety Processing, began rolling out to API customers on Sept. 22, 2026. The second, Private Inference, is expected in preview this fall.

How It Works

Under Zero Data Retention, OpenAI does not retain customer prompts or model responses after a request is processed. Customer content is not available to OpenAI personnel, and enterprise data is not used to train models unless the customer explicitly opts in.

Private Safety Processing allows OpenAI to run automated safety checks without storing customer content. Encrypted data is decrypted inside a hardware-attested safety runtime that blocks human access. Only bounded safety signals and operational metadata exit the protected environment in plaintext.

Customers have two storage options: keep data entirely within their own infrastructure, or store it encrypted on OpenAI's servers using encryption keys the customer controls. OpenAI sets a 30-day time-to-live on encrypted retained safety records.

Private Inference goes further. OpenAI will run model inference on confidential computing hardware that seals data inside a locked enclave that even the cloud operator cannot open. For hospitals, law firms, and banks, that distinction determines whether AI can legally touch client data at all.

The Business Problem OpenAI Is Solving

Enterprise adoption of AI has stalled in some sectors over data control fears. Reporting from The Information showed companies including Palantir, Nvidia, and Booz Allen Hamilton restricting or reconsidering their use of advanced models over concerns that sensitive information could benefit the AI companies providing them.

A controversy around OpenAI's recent Navier-Stokes mathematics breakthrough also raised questions among researchers about whether unpublished work entered into AI systems could influence future models. Telecommunications operator C Spire holds agreements with both OpenAI and Anthropic that bar model training on its data while still permitting collection of technical usage data, illustrating how complicated the line between customer content and metadata can remain even under enterprise contracts.

Sam Altman framed the initiative at DevDay: "Zero Data Retention with private safety processing delivers frontier model safety without storing your content on our servers at all."

Early Partners Signal Healthcare and Enterprise AI Confidence

Glean CISO Sunil Agrawal, whose company was an early development partner, said: "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."

Zach Powers, CISO at healthcare AI firm Abridge, pointed to the collaboration with OpenAI's product, engineering, policy, and leadership teams as central to building trust in a sector where data sensitivity is non-negotiable.

A Direct Shot at Anthropic

The launch puts OpenAI and Anthropic on opposite sides of a key enterprise question. Anthropic moved in June 2026 to require a mandatory 30-day data retention policy for business customers using its most powerful models. OpenAI is now positioning zero-retention as a competitive advantage.

Private Safety Processing and Private Inference are exclusively available to enterprise and API customers. Consumer tiers of ChatGPT, including Free, Plus, Go, and Pro plans, are not covered under Zero Data Retention features.

This article was written by an AI agent. Spotted an error? Send a correction and we will fix it.