Survey: 84% of Organizations Reported at Least One AI-Related Incident in the Past Year — Campus Technology

The research, which surveyed 1,000 senior finance and finance technology leaders across the United States and the United Kingdom, paints a portrait of an industry caught in a high-stakes transition. With 879 of those respondents confirming that their organizations now deploy AI within mission-critical business applications, the prevalence of "AI incidents"—ranging from anomalous financial transactions to untraceable system changes—has reached a level that can no longer be ignored by corporate boards or regulatory bodies.

The Chronology of an AI Governance Gap

The rapid proliferation of AI agents in the workplace did not occur overnight, yet the governance mechanisms intended to oversee them have remained largely stagnant. Historically, enterprise software environments relied on static role-based access controls (RBAC). As organizations began transitioning to cloud-native architectures around 2020-2021, the complexity of managing permissions increased significantly.

Survey: 84% of Organizations Reported at Least One AI-Related Incident in the Past Year -- Campus Technology

By 2023, the emergence of generative AI and autonomous agents shifted the paradigm again. Organizations began moving from human-in-the-loop systems to autonomous agents capable of initiating transactions, modifying user permissions, and interacting with third-party APIs without direct human supervision. The current data indicates that the 2024-2025 period represented a "wild west" phase, where deployment speed took precedence over security architecture. Now, in 2026, the cumulative effect of these unmonitored deployments is manifesting in widespread audit failures and operational disruptions.

Supporting Data: The Scale of the Oversight Crisis

The Delinea report provides granular evidence of why these incidents are occurring with such frequency. While 35% of surveyed organizations have already deployed AI agents empowered to perform actions within core business systems, the security infrastructure has failed to keep pace. Specifically:

  • The Governance Deficit: Only 11% of organizations currently using AI believe their internal governance and security controls are sufficient to keep pace with their deployment speed.
  • Visibility Black Holes: A significant portion of respondents reported a complete lack of centralized visibility. Many agents remain "orphaned"—still operating within business systems even after the employees who originally deployed them have left the company.
  • Monitoring Failures: Less than half (46%) of organizations claim to continuously monitor and manage cumulative access risks across their application stack. This creates a fertile environment for "segregation-of-duties" conflicts, where an AI agent might inadvertently—or maliciously—be granted conflicting permissions across disparate financial and operational systems.
  • Compliance Drift: Perhaps most concerning for shareholders is the finding that 60% of organizations have seen an increase in the number of critical applications operating outside the purview of standard access governance or Sarbanes-Oxley (SOX) compliance requirements over the last two years.

The Paradox of Audit Confidence

One of the most paradoxical findings of the study is the disconnect between the reality of operational incidents and the perceived readiness for external audits. Despite 84% of organizations suffering at least one AI-related incident, 95% of the same cohort expressed confidence that their current access controls would withstand scrutiny during an external audit.

Survey: 84% of Organizations Reported at Least One AI-Related Incident in the Past Year -- Campus Technology

This misplaced confidence may stem from a reliance on legacy compliance frameworks that were never designed for autonomous, machine-based actors. The data suggests that this optimism is unfounded: 26% of respondents admitted to dealing with the same access-related audit findings in consecutive years, and 25% have been unable to resolve specific audit findings for three years or more. These figures suggest that organizations are trapped in a cycle of "check-the-box" compliance, failing to address the structural weaknesses that AI agents are now actively exposing.

Official Perspectives and Industry Implications

The implications for enterprise leadership are profound. Frank Vukovits, chief security scientist at Delinea, notes that the shift toward AI does not absolve the organization of its fiduciary or regulatory responsibilities. "When an agent can approve a transaction or change permissions, the business owns the outcome," Vukovits stated. "You need to know what every agent can do, who owns it, and who approved it, before an auditor asks or your financials are misstated."

From an analytical standpoint, the challenge is not merely technical; it is one of identity management. In traditional IT environments, an identity is tied to a human user with a defined role. With AI, an agent’s "identity" is fluid and often spans multiple systems, making it difficult to define the scope of its authority. Without a single system of record for application ownership—which only 22% of surveyed organizations currently maintain—the "who" behind an action becomes obscured.

Survey: 84% of Organizations Reported at Least One AI-Related Incident in the Past Year -- Campus Technology

Broader Impact: Financial and Regulatory Risks

The failure to govern AI agents effectively carries significant downstream risks. For publicly traded companies, the lack of oversight on systems capable of executing financial transactions could lead to material weaknesses in internal controls over financial reporting (ICFR). Under the Sarbanes-Oxley Act, such weaknesses require disclosure, which can lead to negative market reactions and increased regulatory scrutiny.

Furthermore, the rise of "shadow AI"—applications and agents deployed by business units without the knowledge of the IT or security departments—compounds these risks. When an AI agent operates outside of established governance, it effectively bypasses the audit trails that regulators rely upon to verify that corporate funds and data are being handled appropriately.

Strategic Recommendations for the Enterprise

To bridge the governance gap, industry analysts suggest that organizations must pivot toward an "Identity-First" security model. This involves several critical steps:

Survey: 84% of Organizations Reported at Least One AI-Related Incident in the Past Year -- Campus Technology
  1. Centralized Inventory: Establishing a definitive, automated register of all AI agents, their purpose, their permissions, and their human "owner."
  2. Continuous Access Governance: Moving away from periodic access reviews toward continuous, real-time monitoring of machine-to-machine permissions.
  3. Segregation-of-Duties (SoD) Automation: Using AI-powered security tools to audit other AI agents, ensuring that no single entity—human or machine—possesses excessive or conflicting permissions that could lead to fraud or systemic error.
  4. Formalizing Ownership: Implementing strict policies that require every AI deployment to have a designated human owner who is accountable for the agent’s actions and the security posture of the systems it accesses.

As organizations continue to integrate AI into their core operations, the transition from experimental deployment to enterprise-grade governance is no longer optional. The data provided by Delinea suggests that the "grace period" for AI adoption—where organizations could experiment with minimal oversight—is coming to a close. The future of enterprise AI will be defined by those who can successfully balance the productivity gains of autonomous agents with the absolute necessity of rigorous, verifiable, and transparent governance. Failure to do so may result in more than just audit findings; it could lead to the erosion of trust in the very systems that are intended to power the next generation of business success.