I would like to verify which privacy and model-training controls apply before using private project files in a Codex cloud environment. Does the ChatGPT account-level training opt-out cover Codex tasks, repository contents, uploaded files, tool outputs…