OpenAI Launches "Private Safety Processing" Feature, Targeting Zero Data Retention to Compete with Anthropic

According to Deep Tide TechFlow on August 20, as reported by TechCrunch, OpenAI recently previewed a new feature called "Private Safety Processing" to select enterprise clients. The feature aims to enable automatic cross-session monitoring of potential misuse without retaining any user data. The system is an upgrade based on the existing Zero Data Retention (ZDR) mechanism, capable of analyzing inputs and outputs across multiple conversation sessions to identify distributed malicious behaviors (such as generating malicious code through batch requests to evade detection). Once an alert is triggered, the system sends a "narrow-scope signal" to OpenAI, allowing the platform to decide whether to take enforcement actions, all without requiring manual review of user conversation content.