A chatbot may begin as a simple Q&A service but later schedule vaccine appointments, collect personal information, and require stronger privacy and security safeguards. section 3.3.1
Privacy means protecting and appropriately handling information about identifiable people. For conversational AI, treat submitted data as potentially confidential and choose systems that fit its handling requirements. section 7.1.2
Evaluate privacy before using an LLM to augment conversational AI:
- Check whether the provider may retain submitted data or train on it later. section 7.1.2
- Treat confidential corporate data as unsuitable for any LLM that lacks acceptable handling terms, and involve legal reviewers when needed. section 7.1.2
- Reassess privacy when expanding from informational answers to workflows connected to scheduling systems and databases. section 3.3.1
