OpenAI unveils Private Intelligence: zero data retention with private safety processing
OpenAI's new Private Intelligence initiative pairs Zero Data Retention with private safety processing — available now — with a Private Inference preview arriving this fall.
What happened
OpenAI CEO Sam Altman used the DevDay 2026 keynote at Fort Mason, San Francisco on September 29, 2026 to introduce Private Intelligence, an enterprise privacy initiative that gives customers stronger controls over their data while keeping OpenAI's frontier model safety checks in place.
Two components, two timelines
Private Intelligence has two parts with different schedules:
- Zero Data Retention (ZDR) with Private Safety Processing (PSP) is available now. It allows automated safety reviews of customer content without letting OpenAI personnel access the underlying content.
- Private Inference is in preview and is expected this fall. OpenAI says it combines confidential computing with strict, verifiable controls.
How the safety processing works
Under ZDR with Private Safety Processing, customer content selected through a safety classifier or an approved sampling policy is encrypted and placed into customer-controlled storage — an AWS S3 bucket, Azure Blob container, or Google Cloud Storage — with customer-managed key authorization. The encrypted records are retrieved for automated review inside what OpenAI describes as a "hardware-attested safety runtime" that only predefined safety signals may leave in plaintext. Per OpenAI's documentation, protected records carry a 30-day time-to-live for the automated safety-review system. OpenAI says material processed through PSP cannot be used for model training or shared with other OpenAI groups or partners.
Why it matters
The announcement lands as enterprise concerns about AI data leakage become concrete. The Information reported this month that Palantir, Nvidia, and Booz Allen Hamilton began restricting model use or seeking stronger contractual guarantees over fears that sensitive information could ultimately benefit AI providers. Separately, OpenAI's recent claim that its agents solved the longstanding Navier-Stokes existence and smoothness problem sparked debate among researchers — including NYU mathematician Tristan Buckmaster and Anthropic researcher Levent Alpöge — about whether unpublished work entered into AI systems could influence future models. Altman said OpenAI has been working closely with some of its largest customers, including Cisco, Databricks, and Snowflake, to design privacy and safety systems that meet their standards.
What Altman said
"[Zero Data Retention] with private safety processing delivers frontier model safety without storing your content on our servers at all, and with private inference, we can even provide privacy at the time of inference," Altman said. "Together, we think these set a new standard for privacy and frontier AI."
Sources
- VentureBeat: https://venturebeat.com/security/openais-new-private-intelligence-targets-a-growing-enterprise-concern-who-can-see-your-ai-data
- Superpower Daily: https://superpowerdaily.com/posts/openai-introduces-private-intelligence-to-limit-staff-access-to-customer-ai-data
- CNBC live coverage (via Europe Says): https://www.europesays.com/ie/712119/
Verified September 29, 2026 — facts checked against the sources above on today's date.