Cequence: AI Agent Security Gaps & Data Breach Risks

New research shows a major gap in how companies manage AI agents. While most IT and security leaders are confident their AI agents stay within access limits, only a third have actually implemented those limits effectively. Here's the thing: nearly two-thirds of those surveyed reported AI agents acting beyond their intended scope. This led to serious problems for 29% of them, including data breaches and financial losses. What's more, only 32% of organizations can quickly fix these issues automatically; most rely on slow, manual processes. The study also found that governance isn't keeping up with the rapid deployment of AI. Many organizations are running both generative and agentic AI, and 92% report a surge in AI traffic targeting customer-facing applications. A concerning 31% of AI agent projects are abandoned, often leaving sensitive credentials exposed. The bottom line is that many businesses are deploying AI without proper security guardrails, creating significant risks for their operations and data.