OpenAI Private Intelligence

OpenAI Private Intelligence is an enterprise privacy offering introduced by OpenAI in late October 2026 under the slogan "Frontier intelligence. Private by design." It bundles controls that let companies use OpenAI's most capable models while limiting who — including OpenAI — can see their data.

The two headline pieces are Zero Data Retention (ZDR) for eligible API workflows and Private Inference, a preview option based on confidential computing with strict, verifiable controls. Around them sit access restrictions, approval flows, agent permissions, encryption, customer data ownership and compliance support, plus eligible ChatGPT Work controls.

Key points

1. Zero Data Retention: eligible API requests are processed without being stored.
2. Private Inference (preview): inference runs inside confidential-computing environments designed so that data stays protected even from the operator, with verifiable guarantees.
3. Access & approvals: enterprises can restrict which people, apps and agents can reach data, and require approvals.
4. Agent permissions: as agents take actions (connectors, automations), admins control what they may access or do; ChatGPT Work adds role-based access and app-connection controls.
5. Ownership & compliance: customers own inputs/outputs; encryption and compliance support are part of the package.

Why it matters

Data-access concerns are one of the biggest blockers for enterprise and government adoption of frontier models — especially for agents that touch internal systems. Private Intelligence is OpenAI's answer to on-prem / sovereign-AI pressure from open-weight rivals.

How to get it

1. Read the Private Intelligence page and the "Privacy by product" breakdown.
2. Check whether your API workflows qualify for Zero Data Retention.
3. For Private Inference (preview), register interest or contact OpenAI sales.
4. In ChatGPT Work, configure app connections, automation permissions and role-based access to match your policies.