Enterprise AI app data leaks jump 93%, sparking DLP violations

The amount of sensitive enterprise data uploaded to AI applications has nearly doubled, with a 93% increase in data transfers reported by Zscaler in their 2026 AI Threat Report. Key findings include that over half of the data transfers were attributed to tools like Grammarly and ChatGPT, leading to significant Data Loss Prevention (DLP) violations. Employees often upload sensitive data out of a desire for efficiency, but this poses serious privacy risks.

Zscaler recommends organizations implement measures such as cataloging AI apps, disabling risky defaults, applying zero trust principles, and enforcing inline inspections to mitigate these risks.