ChatGPT introduces per-conversation opt-out for training data

AI Impact Summary

ChatGPT now lets users disable chat history and selectively allow conversations for model training, directly reducing the data pool and enforcing per-conversation consent. This changes how data is ingested into the training pipeline and requires the system to honor user preferences at the data collection stage, with privacy compliance implications and auditability. Engineers should evaluate how to handle opt-out signals, adjust labeling and dataset composition, and plan for alternative data sources to maintain model quality.