The full answer
Privacy starts in design. Creuto maps what data the AI feature needs, where it comes from and who may see the result. Personal or sensitive fields that the model does not need are removed or masked before a request leaves your system.
Retrieval systems respect existing permissions. If a user cannot open a document in your intranet, the AI assistant will not use that document to answer their question. This is enforced at the retrieval layer, not left to the model.
Requests and responses are logged for auditing and quality review, with retention and access controls agreed with you. Logs can be redacted where they would otherwise store sensitive content.
For regulated work, AI features are designed with the relevant rules in mind, such as GDPR, HIPAA, UAE PDPL or India's DPDP Act, depending on your users and sector. Where data must not leave your environment, a private or self-hosted model is an option.
Security testing, including vulnerability assessment and penetration testing before major releases, covers AI features as well, with specific checks for prompt injection and data leakage.