12 AI prompts that leak enterprise data—and how to fix them

**Executive summary:** This report explains the visibility gap and data-exposure risks created by employee use of public generative AI interfaces, enumerates twelve routine leakage scenarios (e.g., contract summarization, code debugging, credential paste, downstream output leakage), and recommends a calibrated technical playbook—inline DLP, redaction, browser isolation, phased discovery-to-enforcement governance—to reduce accidental disclosure of sensitive corporate data.