As enterprises move from experimenting with AI to deploying agentic systems across core business functions, identity has become the limiting factor. AI agents read data, trigger actions, and interact with systems in ways that resemble users and applications, but they do not fit cleanly into existing identity models.
Veza’s new AI Agent Security product is designed to close that gap by extending identity security and governance directly to AI agents.
Announced at the Gartner Identity & Access Management Summit 2025, AI Agent Security is Veza’s first purpose-built offering for AI Security Posture Management (AI SPM).
Rich Dandliker, Chief Strategy Officer at Veza told ChannelE2E that AI SPM is an emerging industry category focused on helping organizations understand what AI exists in their environments, what data and infrastructure it connects to, and how to reduce the associated risk. “AI promises to be a transformation technology,” Dandliker said, pointing out that organizations are rapidly deploying AI agents while often prioritizing speed and ease of deployment over security.
Why IAM, IGA, and NHI tools fall short for AI agents
Most enterprises already run a mix of IAM, IGA, and non-human identity tools, but those platforms were not built with AI agents in mind. IAM typically handles authentication for human users, IGA focuses on HR-driven governance workflows, and NHI tools manage service accounts, secrets, and workloads.
AI agents behave differently from all three. Dandliker explained, some agents run in delegated mode and inherit human permissions, others operate independently using service accounts and API keys, and some behave almost like end users, running locally in browsers or on endpoints. “AI agent identities share bits and pieces with other types of identities, but they are distinct,” Dandliker said, adding that platform providers are creating new constructs that require a new approach to securing agent identities.
Veza positions AI Agent Security as that missing layer. Built specifically for AI agents, it gives security teams a way to apply least-privilege principles and governance consistently, without forcing agents into identity models that were never designed for them.
Using the Access Graph to make AI risk visible
At the center of AI Agent Security is Veza’s Access Graph, which has been extended to include AI agent identities alongside human and non-human identities. Veza pioneered the Access Graph to show true, effective permissions across systems, not just group or role membership.
Dandliker highlighted that the Access Graph provides connectivity across on-prem and cloud systems, visibility into permissions down to the actions that can be taken on specific resources, and a standardized way to translate complex permission structures into language that business users can understand. By layering in activity data, it also helps teams identify unused or unnecessary access and automate cleanup.
Applied to AI environments, this model enables automated discovery of AI agents, clear views into what data and systems each agent can access, traceability to the underlying models in use, and ownership mapping that ties every AI agent back to a responsible human. This gives security teams a concrete way to assess AI risk instead of treating it as an abstract problem.
Turning AI agent security into action
Veza emphasizes that AI Agent Security is designed to fit into existing workflows rather than operate as a standalone alerting tool. The platform integrates with common identity providers such as Microsoft Entra ID, Active Directory, and Okta, tying agent ownership directly to human identities.
This linkage allows ownership to follow standard joiner, mover, and leaver processes. If a human owner leaves the organization and their account is deactivated, ownership of their AI agents can automatically be reassigned to an active user, reducing orphaned or unmanaged agents.
On the governance side, Veza positions AI Agent Security as a foundation for agent-focused identity governance. Unlike legacy IGA tools that often rely on manual inventories, Veza automatically discovers agents and maps their permissions, giving reviewers the context needed to make informed access decisions. Reviews can trigger actions in downstream systems like ServiceNow or directly remove rejected permissions within the platform.
For incident response, Veza provides risk scoring and blast-radius analysis when an AI agent is compromised. Dandliker highlighted that this context can be pulled into existing SOAR platforms or identity protection tools through APIs and prebuilt integrations, helping teams prioritize response based on actual impact.
What this means for MSPs
For MSPs, Veza is clear about the current state of the platform. Dandliker said that while unified multi-tenant management is not yet available, the platform offers a broad API surface that MSPs can use to programmatically manage visibility and workflows across individual customer tenants. This reflects a broader industry reality as service providers adapt operating models to account for agentic AI.
By introducing AI Agent Security, Veza is formalizing AI SPM as a distinct discipline rooted in identity. As AI agents become embedded in security operations, development pipelines, and customer-facing workflows, visibility, ownership, and access control will determine whether those deployments remain manageable. Veza’s approach treats AI agents as first-class identities, governed alongside humans and machines on a single, unified identity security platform.
