TechFlow reports, August 20. According to TechCrunch, OpenAI recently previewed a new feature named "Private Safety Processing" to select enterprise clients, aiming to enable automatic cross-session monitoring of potential abuse without retaining any user data.
The system is upgraded based on the existing Zero Data Retention (ZDR) mechanism, capable of analyzing inputs and outputs across multiple conversation sessions to identify distributed malicious activities (such as malicious code generation via batched requests to evade detection). Once a warning is triggered, the system will send "narrow-scope signals" to OpenAI, and the platform will decide whether to take enforcement measures, with no manual review of user conversation content throughout the process.





