AI Agents Spark Cybersecurity Incidents at Two Thirds of Companies, CSA Report Finds

Artificial intelligence agents are rapidly becoming a staple in enterprise networks, but their unchecked deployment is causing serious trouble. According to a new report from the Cloud Security Alliance (CSA), conducted in partnership with Token Security, two thirds of organizations have suffered from AI agents cybersecurity incidents over the past year. These incidents have led to data exposure, operational disruptions, and financial losses, raising urgent questions about governance and oversight.

The report, titled Autonomous but Not Controlled: AI Agent Incidents Now Common in Enterprises, published on April 21, warns that most organizations lack a formal strategy for decommissioning AI agents. This oversight leaves them vulnerable to ongoing risks. As companies race to adopt AI, the gap between deployment and security is widening.

The Visibility Gap: Known vs. Unknown AI Agents

One of the most striking findings is the disconnect between perceived and actual visibility. While 68% of respondents expressed high confidence in their ability to track AI agents on their networks, 82% admitted to discovering previously unknown agents in the past year. This paradox highlights a critical blind spot.

Internal automation environments and large language model (LLM) platforms were the most common hiding spots for these rogue agents. The CSA report notes, “This gap highlights a distinction between operational visibility and complete governance assurance, limiting the effectiveness of control models that depend on known and bounded agents.”

When cybersecurity and infrastructure teams are unaware of AI agents deployed by employees, securing the network becomes nearly impossible. This lack of awareness has directly contributed to the rise in AI agents cybersecurity incidents.

Consequences: Data Exposure, Disruptions, and Financial Hits

The operational fallout from these incidents is significant. Among the 65% of organizations that experienced at least one incident, the most common consequences included data exposure (61%), operational disruption (43%), and unintended actions in business processes (41%).

Financial losses were reported by 35% of affected firms, while 31% faced delays in customer-facing or internal services. The paper warns that AI agent incidents are now hitting core enterprise functions, from data protection to service delivery. As the report states, “For organizations, this shifts AI agent governance from a technical oversight issue to a business risk management concern.”

Why Financial and Operational Risks Are Rising

Building on this, the report emphasizes that AI agent behavior must be integrated into broader security, compliance, and operational resilience strategies. Treating it as an isolated automation challenge is no longer viable. Companies must perform thorough risk assessments to apply appropriate controls.

The Decommissioning Problem: Forgotten Agents Pose Persistent Threats

Governance around AI agent decommissioning is particularly weak. Only one in five organizations have formal processes for retiring AI agents. As a result, many agents persist on networks long after their purpose is fulfilled.

These forgotten agents often retain credentials, permissions, or operational hooks. This creates a ticking time bomb for cybersecurity. The CSA warns that as more AI agents become part of enterprise networks, the problem of agent sprawl will only amplify risks. Without proper end-of-life governance, AI agents cybersecurity incidents will likely increase.

How to Strengthen AI Agent Security and Governance

In response to these challenges, the CSA has issued a set of actionable recommendations for organizations. Hillary Baron, assistant vice president of research at the CSA, explains, “AI agent security and governance encompass an interconnected system spanning visibility, lifecycle management, policy, and monitoring. While foundational controls are in place, gaps in consistency and end-of-life management remain.”

To address these gaps, the CSA advises firms to:

• Maintain visibility across AI agents — Ensure agents operating across SaaS platforms, internal systems, and LLM environments are identified and within governance scope.
• Define and document agent purpose — Establish intended functions to set operational boundaries and align access accordingly.
• Apply lifecycle governance consistently — Extend onboarding, ownership, review, and decommissioning processes across the full agent lifecycle.
• Evaluate actions based on risk and authorization — Use contextual signals such as action risk and explicit human approval to guide decision-making.
• Align monitoring with agent activity — Evolve from periodic oversight toward more continuous or event-driven detection models.
• Incorporate agents into enterprise risk models — Treat AI agents as part of broader security, compliance, and operational resilience frameworks.

For more insights on managing AI risks, check out our guide on AI security best practices. Additionally, learn about cloud security strategies to protect your digital assets.

As AI agents gain greater autonomy, governance must evolve into a more unified, operational model. The stakes are high, but with proactive measures, organizations can harness the power of AI without falling victim to its risks.