The AI provider's privacy policy in effect when the tool was used reportedly allowed disclosure of prompts and outputs to government authorities, so there was no reasonable expectation of confidentiality.