OpenAI Simulates Deployments to Catch Model Risks

OpenAI published Deployment Simulation, a pre-deployment safety method that replays past conversations through a new candidate model to estimate undesired behavior frequency. The approach targets non-tail risks and has already informed mitigations and deployment decisions. It cannot measure behaviors occurring less than once in 200,000 messages.