OpenAI previews Private Safety Processing to keep zero data retention

OpenAI announced Private Safety Processing, a zero-data-retention (ZDR) option that lets eligible enterprise API customers keep prompts and model responses on their own or encrypted OpenAI infrastructure, with staff never seeing the content. The system flags multi-interaction misuse patterns—such as coordinated probing or agent drift—while still preventing the data from being used for model training unless the customer opts in.