Prompts can contain personal concerns, confidential work, source code, research, legal material, account details, or information about other people. Privacy depends on collection, encryption, access control, provider use, model training policies, logging, retention, deletion, and disclosures to tools or subprocessors.
Linking prompts with verified identity increases the sensitivity of the record because content and person can be analyzed together. Services can reduce risk through data minimization, separation of identity from content, short retention, local processing, privacy-preserving credentials, and clear user controls.
ELI5
Prompt privacy is about protecting what people type, upload, or ask an AI system. A prompt can reveal much more than an ordinary search because it may contain private problems, documents, or work.
For example, a whistleblower might ask for help understanding a document. Keeping the prompt separate from verified identity and deleting unnecessary logs reduces the chance that the question exposes the person.
