A new report highlights a significant blind spot in enterprise security architectures concerning the proliferation of third-party AI agents. The forthcoming 2026 State of Agent Security Report indicates that current security models, often designed to protect AI systems directly chosen and deployed by an organization, frequently fail to account for AI capabilities embedded within third-party products that operate outside the purview of traditional identity and access management (IAM) infrastructure.
The report's findings reveal that a substantial number of third-party products now incorporate AI functionalities. Specifically, approximately 1,280 such products were identified as embedding AI in the environments studied. A notable portion of these, around 282, are integrated with enterprise single sign-on (SSO) systems. This integration typically means their access and user authentication can be managed and monitored through existing identity infrastructure, providing a degree of visibility and control.
However, the vast majority of these AI-enabled third-party products, roughly 1,000 according to the report, operate without authenticating through the organization's primary identity stack. This lack of integration means they are effectively "invisible" to security teams relying on IAM systems to govern access and enforce policies. The issue is not one of deliberate concealment, but rather a fundamental limitation: an identity infrastructure can only manage and secure what authenticates through it.
This gap presents a considerable challenge for enterprise security. Third-party agents, even those embedded in seemingly innocuous applications, can access, process, and potentially exfiltrate sensitive data. Without proper authentication and authorization mechanisms tied to the enterprise's identity infrastructure, these agents operate with an unknown security posture and an unmanaged risk profile. This class of issue typically means that traditional security controls like multi-factor authentication, session monitoring, and access revocation may not apply.
Mitigation for this class of problem commonly involves a comprehensive discovery process to identify all third-party applications and services in use, followed by a thorough assessment of their embedded AI capabilities and data access requirements. Organizations are often advised to implement robust third-party risk management programs, including contractual agreements that mandate specific security controls and audit rights. Technical solutions might include network segmentation, API security gateways, and specialized agent security platforms designed to monitor and control the behavior of AI agents regardless of their authentication path.
The emergence of AI as an embedded feature across a wide array of third-party products underscores a broader trend in enterprise technology: the decentralization of advanced capabilities. As AI becomes a ubiquitous component rather than a standalone system, security strategies must evolve beyond protecting explicitly deployed AI models to encompass the pervasive and often unmanaged AI agents operating within the digital supply chain. This shift necessitates a re-evaluation of how organizations discover, assess, and secure all forms of intelligent automation interacting with their data and systems.
