OpenAI launches privacy-focused AI misuse detection system

OpenAI announced a new privacy-focused safety system on August 19, 2026, designed to detect potential misuse of its artificial intelligence models across multiple conversations without retaining any customer data. The system, called Private Safety Processing, is being previewed with early customers and is expected to roll out more widely in September alongside a technical white paper.

The development represents an enhancement to OpenAI's existing Zero Data Retention policy for eligible API customers, which already ensures prompts and model responses are not stored after a request is processed. Private Safety Processing extends this principle by using automated systems to analyze patterns across related interactions, rather than evaluating each session in isolation. This approach aims to identify more complex, long-horizon risks that might be missed when looking at single prompt-response pairs.

"We're seeing with more capable frontier models that often, risks are emerging not just by looking at one single prompt and response pair, but when you look over time at multiple interactions," said Aleah Houze, OpenAI's head of product policy, at a press briefing. She provided an example...