Particle.news

Enterprises Lack Visibility and Controls as AI Traffic Outpaces Security

The shortfall means firms can often detect AI problems but lack the architecture to stop data leaks, block malicious prompts, or filter unsafe outputs in real time.

Overview

  • Check Point’s late-May 2026 Cloud Security Report, based on 1,042 IT and security pros, found 77% of organisations updated AI security strategy but only 26% have an architecture that can enforce those policies.
  • Only 5% of firms report full visibility into which AI tools employees use and where data flows, and 78% have either confirmed or suspect AI-related incidents such as shadow AI use, AI-driven phishing, or sensitive data leaks.
  • Network and inspection tools struggle with API-heavy AI traffic: only 24% can fully inspect AI traffic without harming performance and 67% report fragmented policies across hybrid environments.
  • Prevention controls are scarce at key enforcement points: 13% can block malicious prompts, 16% can stop sensitive data reaching AI services, 5% can block unsafe AI outputs, and only 17% broadly deploy runtime LLM controls.
  • Report authors urge an architecture-first fix that inventories AI assets, governs access, installs prevention at runtime, centralises policy authority, and unifies hybrid security to reduce agent-driven and data-exfiltration risks.