OpenAI bundles its business privacy protections as 'Private Intelligence'

OpenAI on Oct. 30 launched Private Intelligence, a set of privacy protections for businesses that use its AI on sensitive information. OpenAI announced it in a post on X that day, the Norwegian tech site hogby.ai reported, and describes it on its website.

When a business sends a question to an AI model online, the text travels to the AI company's computers to be processed. OpenAI says businesses own what they type and what the AI sends back, and that it strictly limits its own access.

The package includes Zero Data Retention, in which OpenAI does not keep the business's questions and answers for eligible uses. The new piece, Private Inference, uses confidential computing, special hardware that keeps data sealed off even from the people who run the computers, OpenAI says. It is still a preview for software developers. OpenAI's own technical paper says Private Inference for now covers only automated safety checks, not the whole request, hogby.ai reported. Businesses can also limit what AI agents, programs that act on their own, may do and when they need approval, OpenAI says.

Is it for everyone?

Not for now. OpenAI's page is pitched at "enterprises" and "business data," and says Zero Data Retention and Private Inference are offered only for "eligible workflows." Private Inference is labeled an API preview, meaning it is for companies building their own software on OpenAI's paid service for developers. OpenAI's white paper limits it to customers that meet eligibility and capacity requirements, hogby.ai reported. OpenAI has not announced a version for people using ChatGPT on their own, and it has not published prices. Businesses can fill out an interest form.

What happens when you use regular ChatGPT

If you use ChatGPT's Free, Plus or Pro plans, OpenAI says it may use your conversations to train its models, unless you opt out. To opt out, go to Settings, then Data controls, and turn off "Improve the model for everyone." That stops OpenAI from training on new conversations. But if you click thumbs up or thumbs down on an answer, OpenAI may use that entire conversation for training even after you opt out. OpenAI says it works to strip personal details from training data and does not sell user data.

Business accounts get different rules. By default, OpenAI does not train on conversations from ChatGPT Business, Enterprise or Edu, or on data sent through its developer service.

A Temporary Chat is not used for training and does not appear in your history, but OpenAI may keep a copy for up to 30 days for safety reasons. When you delete a regular chat, OpenAI says it schedules it for permanent deletion within 30 days, unless it must keep it longer for security or legal reasons.

Is it risky to share private information?

There is some risk. OpenAI's privacy policy warns that "no Internet or email transmission is ever fully secure" and that users "should take special care in deciding what information you provide." Saved chats stay on OpenAI's servers until you delete them. OpenAI may share data with government authorities when the law requires it, such as under a subpoena, and administrators of a work account can see your conversations, the policy says.

Legal fights can also keep chats around longer. In 2025, a court order in The New York Times' copyright lawsuit against OpenAI forced the company to keep consumer ChatGPT conversations, even deleted ones. OpenAI says that order ended Sept. 26, 2025, but it is still holding some April to September 2025 user data under lock for the case.

Companies have learned this the hard way. In 2023, Samsung temporarily banned employees from using ChatGPT on company devices after sensitive internal data was accidentally leaked to the chatbot, Bloomberg and TechCrunch reported. Samsung said such data is hard to retrieve and delete once it is on outside servers.

The safest course, based on OpenAI's own warning, is to leave out passwords, account numbers, medical details and work secrets, or to use a work account your employer has approved.

Source: OpenAI, Private Intelligence page, hogby.ai, Oct. 31, 2026, OpenAI Help Center on model training, Temporary Chat and deleting chats, OpenAI US privacy policy, OpenAI enterprise privacy, OpenAI on the New York Times data order and TechCrunch, May 2, 2023.