AI Security Threats Loom as Enterprise Usage Jumps 91%

Zscaler’s ThreatLabz 2026 AI Security Report analyzes 989.3 billion AI/ML transactions from 2025, finding a 91% surge in enterprise AI usage across 3,400+ apps, heavy concentration in the U.S., and OpenAI as the top vendor. The report warns of systemic risk: critical vulnerabilities were observed across all AI systems reviewed, with 90% compromised in under 90 minutes and a median time to first critical failure of 16 minutes. Enterprises transferred 18,033 TB of data to AI/ML apps, with massive flows to Grammarly and ChatGPT, and 410 million DLP violations tied to ChatGPT alone. The report anticipates agentic AI will increasingly automate the attack lifecycle, accelerating reconnaissance, exploitation, and lateral movement.