A recent report from Tenable has raised significant concerns about a developing phenomenon termed the “AI exposure gap” among organizations. This gap arises from the rapid deployment of artificial intelligence tools and cloud-native services by enterprises, which is occurring at such a pace that security and governance teams struggle to keep up. The findings highlighted in the “Cloud and AI Security Risk Report 2026” reflect a critical urgency for organizations to reassess their security postures in light of evolving technologies.
The report scrutinized telemetry data collected from public cloud environments and enterprise ecosystems, covering the period from April to October 2025, while also incorporating AI-related data collected up until December 2025. The results are alarming, revealing that numerous organizations are inadvertently broadening their attack surfaces. This increased vulnerability manifests in several ways, including the proliferation of overprivileged AI identities, insecure third-party libraries, and exposed cloud credentials.
One particularly troubling statistic from the report indicates that 18% of enterprises possess identities with excessive privileges, which AWS’s artificial intelligence services can assume at alarming rates. These identities, which hold extensive administrative rights, are often not subjected to audits that adhere to the least privilege policy, leading to significant security risks.
Tenable’s Senior Product Marketing Manager for Cloud and Identity, Brinton Taylor, discusses what he refers to as the “Dormant Privilege Gap.” This gap represents a critical vulnerability in organizations’ security frameworks. While many businesses may focus intently on securing human logins through multi-factor authentication (MFA), they often overlook the “keys to the kingdom” held by AI agents and other non-human identities (NHIs).
Taylor elaborated on this issue, stating that 52% of organizations have NHIs with dangerously excessive permissions, which represents a higher risk than the equivalent figures for human identities at 37%. Acknowledging that many firms believe active monitoring alone suffices for security, Taylor pointed out that a staggering 49% of identities classified as critical-severity permissions are actually dormant, remaining unused for periods of 90 days or more. This situation creates an expansive, silent attack surface that malicious actors can potentially exploit.
As organizations race toward deeper AI integration, they may remain blissfully unaware of the risks posed by overprivileged Identity and Access Management (IAM) roles. Taylor emphasized that 18% of organizations have unwittingly created these roles, which AWS AI services can quickly assume, thereby establishing an unguarded pathway for lateral movement by attackers.
The report also pinpointed a rising threat associated with AI—the vulnerabilities stemming from software supply chains. Findings reveal that 70% of companies have integrated AI or Model Context Protocol (MCP) packages from external sources into their infrastructures. Furthermore, 86% of organizations are executing critical code packages provided by third parties, with a troubling 13% of firms adopting packages with confirmed exploit records.
Identity management issues compound these concerns, as over half of the surveyed organizations (53%) have third-party identities capable of assuming permission levels deemed exceedingly high. Consequently, this situation necessitates a broader security strategy that incorporates external parties. Additionally, the report highlights that 65% of firms have high-value resources exposed due to neglected cloud credentials, while 57% are operating with outdated, end-of-life technology.
According to Taylor, the next major security breach is unlikely to arise from sophisticated zero-day exploits, but rather from vulnerabilities that organizations are already aware of yet have failed to adequately address. He remarked, “The next breach will reveal that organizations prioritized velocity over hygiene, leaving a trail of ‘sitting duck’ workloads and unmanaged supply chain dependencies.” As organizations embrace AI tools for active business purposes, it is troubling to note that 55% of them are currently demonstrating this trend without fully acknowledging the “Infrastructure Debt” it incurs.
The fallout from any upcoming security incidents will likely expose two critical failures: “Workload Neglect” and “External Over-trust.” Taylor highlighted that 82% of organizations are currently operating workloads containing known critical vulnerabilities, while 53% have granted external accounts the ability to assume excessively critical permissions, effectively tethering their security to the weakest links in their vendor chain.
When the inevitable exploit occurs, investigations will likely reveal not an advanced new zero-day attack, but instead a “forgotten” credential that remains present in 65% of organizations, facilitating an attacker’s entry.
Ultimately, the report underscores that organizations face a challenge characterized by engineering speeds that outrun governance capabilities. Tenable asserts that many enterprises continue relying on disconnected security solutions and remediation strategies, which inadequately capture the complex interconnections between AI services, cloud workloads, identities, and data. In response, Tenable advocates for the adoption of context-based exposure management practices, emphasizing the importance of recognizing dependencies related to identity, access paths, workload exposure, and inter-cloud connections. This shift could play a crucial role in mitigating risks and developing a more secure technological environment as businesses continue to embrace AI and cloud solutions.