OpenAI Preserves Zero Retention for Frontier Models
A privacy-preserving safety system will analyze multi-step risks without giving OpenAI personnel access to customer content.
It tackles the central enterprise tradeoff between monitoring autonomous frontier models and protecting highly sensitive customer data.