Zscaler: Enterprise AI Systems Can Be Breached In Under 2 Hours
Zscaler, Thursday, January 29th, 2026
As reported by Tech Radar, new research from Zscaler highlights critical security vulnerabilities in enterprise AI systems, warning that many can be breached in less than two hours, leading to rapid data loss.
Zscaler's tests revealed that 90% of enterprise AI systems have critical vulnerabilities discoverable in under 90 minutes, with the median time to first critical failure at just 16 minutes. In extreme cases, defenses were bypassed in a single second.
This susceptibility to rapid exploitation is concerning given the 91% year-on-year increase in AI and machine learning activity across over 3,400 applications. The finance & insurance sector leads AI adoption, but technology and education saw significant usage spikes.
Enterprise data transfers to AI applications surged 93% year-on-year, with tools like Grammarly and ChatGPT becoming major repositories of corporate intelligence. Many organizations lack basic inventories of their AI models, leaving sensitive data exposed.