OpenAI on Friday introduced OpenAI Private Intelligence, which it says “gives businesses stronger controls over their data while using our most capable models”. Its product page offers Zero Data Retention “for eligible API workflows” and Private Inference, a confidential-computing option marked “(preview)”.
The page lists seven protections, by our count. Four of the seven carry a qualifier in OpenAI’s own wording: three include an ‘eligible’ qualifier on at least part of what they offer, and one is marked preview.
Business Engineer Course · The FDE Playbook
An evergreen Business Engineer course on forward-deployed engineering, how AI gets put to work inside organisations. It teaches the general idea only and does not discuss OpenAI or Private Intelligence.
The key insight:OpenAI has bundled its enterprise data protections as OpenAI Private Intelligence and added Private Inference, a confidential-computing option, in preview; by our count four of the seven protections on its page include an eligibility or preview qualifier.
In its post on X at 15:12 UTC on 30 October, OpenAI wrote: “Privacy should advance as fast as intelligence.” It said it is building the protections “in close collaboration with some of our largest customers” to help businesses “apply frontier AI to their most sensitive work”.
The page describes Private Inference as combining “confidential computing with strict controls designed to protect your prompts and responses while the model processes them”. Zero Data Retention, it says, “means we don’t retain your prompts or model responses after processing”.
The seven protections listed on OpenAI’s Private Intelligence page, grouped by the qualifier OpenAI attaches: three with no condition stated, three with an eligibility qualifier on part of what they offer, and one in preview. The grouping is our count.
What Was Posted on X
Privacy should advance as fast as intelligence. OpenAI Private Intelligence gives businesses stronger controls over their data while using our most capable models.
Weโre building these protections in close collaboration with some of our largest customers to help businesses apply frontier AI to their most sensitive work. https://openai.com/business/private-intelligence/
OpenAI announced Private Intelligence on X on 30 October, saying it gives businesses stronger controls over their data while using its most capable models, and that it is building the protections with some of its largest customers.
The Seven Protections
The page groups four areas. On access, OpenAI says: “We set strict limits on our access to business content, and provide ZDR and Private Inference for eligible workflows.” On ownership: “You own your business data, including your inputs and outputs.”
On agents, the page says: “You decide what agents can access, which actions they can take, and when they need your approval.” On security: “We protect your data with encryption, strict access controls, and ongoing threat detection.”
The page then lists three mechanisms: Zero Data Retention, Private Inference (preview), and compliance, which it describes as “SOC 2 Type 2 audits, ISO certifications, HIPAA support through Business Associate Agreements for eligible services, and FedRAMP support for designated government offerings”.
Our Count
Of the seven protections, three are stated on the page with no qualifier: data ownership, agent actions and data security. Three include an ‘eligible’ qualifier on at least part of what they offer. Access restrictions pair “strict limits on our access to business content” with ZDR and Private Inference “for eligible workflows”. Zero Data Retention is “available for eligible API workflows”. Under compliance, HIPAA support is “for eligible services” and FedRAMP support is “for designated government offerings”.
One, Private Inference, is marked preview. That is four of seven with a qualifier, by our count of the page’s own words.
As we read the page, the new element is Private Inference, which OpenAI says protects prompts and responses while the model processes them; Zero Data Retention covers what happens after processing.
As we read the agent section, OpenAI frames control over agent access, actions and approvals as something the business sets, not the model provider.
As we read the announcement, the qualifiers matter: Zero Data Retention is for eligible API workflows and Private Inference is in preview, and the page does not list which workflows qualify.
“You decide what agents can access, which actions they can take, and when they need your approval.”
Three Implications
FOR ENTERPRISE BUYERSOpenAI says Zero Data Retention is available for eligible API workflows and that Private Inference is in preview.
FOR REGULATED INDUSTRIES The page lists SOC 2 Type 2 audits, ISO certifications, HIPAA support for eligible services and FedRAMP support for designated government offerings.
FOR AGENT DEPLOYMENTS The page says businesses decide what agents can access, which actions they can take and when they need approval.
The framework describes one moat this way: “The Container Moat โ built by closed data loops inside a customer’s environment, nascent but the deepest position in the stack.”
The framework is the owner’s general argument about where AI value accumulates, not evidence about OpenAI’s announcement; it does not discuss Private Intelligence.
What Is Not Established
Neither the post nor the page names the customers OpenAI says it is working with. The page gives no prices and no date for Private Inference to leave preview. It does not say which workflows or API endpoints count as eligible for Zero Data Retention. It does not list which models Private Inference supports, beyond the post’s reference to “our most capable models”.
This Week In The AI Supercycle
Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls
This week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.
OpenAI has packaged its enterprise data protections as OpenAI Private Intelligence, adding Private Inference, a confidential-computing option, in preview. By our count of its product page, four of the seven protections it lists include an eligibility or preview qualifier on at least part of what they offer.
A note on sourcing. Product details come from OpenAI’s Private Intelligence product page and its X post of 30 October 2026, read that day. The grouping of the seven protections is our own count of the page’s wording.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
