Employees Enter Sensitive Data Into GenAI Prompts Far Too Often

Harmonic Security analyzed thousands of prompts submitted to major GenAI platforms and found that 8.5% contained sensitive data, predominantly customer (45.77%) and employee (27%) information, with smaller but growing shares involving legal/finance, security data, and code. The article weighs the competitive risks of not adopting GenAI against the exposure risks of using it and recommends moving beyond simple blocking to robust AI governance: real-time monitoring of inputs, plan management, visibility and data classification, enforced workflows, and employee training to mitigate leakage.