OpenAI previews new privacy monitoring system for business AI use

In short: OpenAI says it will test a new way to watch for AI misuse without keeping a customer’s data.

What happened

OpenAI announced a preview of a service called Private Safety Processing for select customers. OpenAI says it is designed for enterprise users, meaning businesses that use AI tools at work.

The idea is to detect misuse, like someone trying to get help making malware for a cyberattack, while retaining none of the customer’s data. “Data retention” is how long a company keeps records of what you typed and what the AI answered. Think of it like whether a customer service call is recorded and saved, or listened to and then immediately discarded.

OpenAI already offers what it calls Zero Data Retention for many API customers. An API is a way for a company’s software to connect to OpenAI’s software (like plugging one app into another). OpenAI says Private Safety Processing expands this by looking for patterns across multiple conversations over time, not just within a single session.

This move comes as OpenAI and Anthropic compete on privacy protections for business customers. Anthropic recently said it may keep data for 30 days for certain “covered models,” which has worried some companies that handle sensitive information.

Why it matters

More businesses are using AI for work that can include private details, like customer messages, legal drafts, or internal plans. These companies want strong safety checks, but they also want assurance that their information will not be stored or reviewed by people. If OpenAI’s approach works as described, it could make it easier for organizations to use powerful AI systems while reducing concerns about where their data ends up.

Source: TechCrunch AI