Sensitive information disclosure means confidential information reaches unauthorised people through an AI, for example via inputs, answers or connected tools.
Example
A chatbot with access to internal documents accidentally tells a customer salaries from a staff list.
For pros
OWASP lists this among the top risks for LLM applications. Causes include overly broad access rights, prompt injection, stored chats or training on customer data. Protection: minimise data, check permissions per user, filter outputs and choose providers with clear data protection commitments.
