Zscaler report reveals 83% surge in enterprise AI transactions, 410 million DLP violations

Zscaler’s latest ThreatLabz 2026 AI Security Report exposes the rapid expansion of AI adoption in enterprises and the security challenges accompanying it. The report, based on nearly a trillion AI/ML transactions from 9,000 organizations in 2025, reveals an 83% year-over-year increase in AI/ML transactions and a 93% surge in data transfers to AI applications—alongside 410 million data loss prevention (DLP) policy violations tied to ChatGPT alone.

Tarun Mottlia

Editor, Lazyfounder

Published Oct 22, 20264 min read

Zscaler report reveals 83% surge in enterprise AI transactions, 410 million DLP violations — Image: Zscaler via source

30 SEC SUMMARY

Zscaler’s ThreatLabz 2026 AI Security Report reveals an 83% year-over-year surge in AI/ML transactions across enterprises, with data transfers to AI applications rising 93% in 2025.
The report analyzed nearly a trillion AI/ML transactions from 9,000 organizations, uncovering 410 million DLP policy violations tied to ChatGPT alone.
AI adoption is occurring outside traditional procurement and security reviews, with 69% of cybersecurity leaders suspecting unauthorized use of public generative AI.
The enterprise AI stack now includes developer tools, AI agents, employee-used services, and embedded AI features, complicating visibility and governance.

KEY HIGHLIGHTS

AI/ML transactions across Zscaler’s platform increased 83% year-over-year, with data transfers to AI/ML applications surging 93% in 2025.
The number of applications generating AI/ML traffic nearly quadrupled to over 3,400.
Zscaler recorded 410 million DLP policy violations tied to ChatGPT, including unauthorized sharing of Social Security numbers, source code, and medical records.
69% of cybersecurity leaders suspect or have evidence of employees using prohibited public generative AI.
AI adoption often bypasses traditional procurement and security review processes, expanding through embedded features and APIs.
Salesforce customers more than doubled their agentic AI workforce between February 2025 and April 2026.

AI adoption outpaces security oversight

Zscaler’s ThreatLabz 2026 AI Security Report analyzed nearly a trillion AI/ML transactions from 9,000 organizations in 2025, revealing an 83% year-over-year increase in AI/ML transactions. The number of applications generating this traffic nearly quadrupled to over 3,400, indicating widespread adoption across enterprises.

Data transfers to AI/ML applications surged by 93% in the same period. However, this growth has come with significant security risks: Zscaler recorded 410 million DLP policy violations tied to ChatGPT alone, including unauthorized sharing of Social Security numbers, source code, and medical records.

The report highlights that AI adoption often occurs outside traditional procurement or security review processes. This "shadow AI" phenomenon complicates governance, as 69% of cybersecurity leaders suspect or have evidence of employees using prohibited public generative AI tools.

The expanding enterprise AI stack

The enterprise AI stack now extends beyond developer tools and employee-used services to include AI agents and embedded AI features in existing platforms. Salesforce customers, for example, more than doubled their agentic AI workforce between February 2025 and April 2026, reflecting this shift.

AI interactions with sensitive data are increasingly happening in areas not actively monitored or classified as AI usage by security teams. This includes APIs connecting AI services to internal systems and AI features built into software platforms, which can expand an organization’s AI footprint discreetly.

The report emphasizes that blocking AI tools tends to drive usage underground rather than reduce it. Instead, visibility into AI usage—tracking tools, data flows, and permissions—is critical for governance and security.

Context: AI governance challenges in recent news

The Zscaler report underscores persistent challenges in AI governance, echoing recent developments in the space. Salesforce’s renaming of its AIForce platform to SIForce earlier this year—following a U.S. executive order encouraging the use of "super intelligence" terminology—reflects ongoing shifts in how enterprises frame and regulate AI tools.

Nvidia’s introduction of the OpenShell Reference Architecture, aimed at ensuring safe deployment of AI agents, aligns with Zscaler’s findings on the need for structured AI security frameworks. Both highlight the tension between rapid AI adoption and the safeguards required to manage risks effectively.

What this means

Lazyfounder analysis — our interpretation, not reported fact.

For enterprises, the Zscaler report is a reminder that AI adoption is not just about deploying cutting-edge tools—it’s about managing the unintended consequences. The surge in DLP violations and shadow AI use shows that security teams are often playing catch-up, reacting to risks rather than anticipating them. The rise of embedded AI features in platforms like Salesforce means that even organizations with strict AI policies may still be exposed to vulnerabilities through indirect usage.

The report also signals that outright bans on AI tools are ineffective. Instead, enterprises need proactive strategies to monitor AI transactions, enforce policies in real time, and integrate Zero Trust principles into their AI governance frameworks. For founders and operators, this means prioritizing visibility and control as part of their AI roadmap—not as an afterthought.

Key takeaways

Enterprises must balance AI adoption with security, as rapid deployment outpaces governance frameworks and increases exposure to data leaks.
Shadow AI—unauthorized or unmonitored AI use—poses significant risks, requiring visibility into tools, data flows, and permissions to enforce policies effectively.
The rise of embedded AI features in existing platforms means security teams may overlook critical interactions with sensitive data, necessitating broader monitoring strategies.
Blocking AI tools may push usage underground, making proactive governance and Zero Trust principles more effective for managing risks.

Sources

The Next Web · 2026-10-22
Your enterprise has more AI than you think, and security can’t see it all

This story is an original summary drafted with AI by Lazyfounder from the reporting listed above and checked by automated validation. Facts are attributed to their original publishers; sections marked as analysis are Lazyfounder's. Where a source is in another language, facts were machine-translated and quotations are reported, not reproduced. Read the original coverage via the links, and see our AI policy and corrections policy.

About the author