OpenAI wants to get better at spotting people abusing its frontier AI models without asking privacy-conscious customers to give up Zero Data Retention. In an August 19 announcement, OpenAI introduced Private Safety Processing, a system designed to detect suspicious behavior across multiple interactions while preserving ZDR for eligible API customers.
OpenAI says its existing ZDR safeguards evaluate requests individually, making patterns spread across several interactions harder to detect. Private Safety Processing is supposed to fill that gap without giving OpenAI employees access to the underlying prompts or responses.
For now, the system is…
Read the full article at DIGITALTRENDS.COM
