AI Agent Identity Consolidation: NHI Market Concentration Risk

Key Takeaways

• On May 4, 2026, Cisco announced its intent to acquire Astrix Security for approximately $400 million, absorbing one of the market’s leading purpose-built non-human identity (NHI) security platforms into a large security infrastructure vendor [1][2].
• The acquisition follows CrowdStrike’s January 2026 announcement of its $740 million acquisition of SGNL, a runtime access enforcement platform [3][13]. These two deals in under five months signal that NHI security has crossed from an emerging category into active platform consolidation.
• According to Astrix’s co-founders, NHIs — API keys, service accounts, OAuth tokens, machine credentials, and increasingly AI agents — outnumber human identities in enterprise environments by as much as 100 to 1, and the company estimated that NHIs remain subject to less than 5% of enterprise identity governance oversight [4][5].
• Cisco’s own AI Readiness Index — a survey produced by the acquiring company — found that only 24% of organizations can control AI agent actions with proper guardrails and monitoring, and just 31% feel fully equipped to secure their agent AI systems [6].
• Platform consolidation accelerates the maturation of NHI security capabilities but introduces new vendor concentration risks: enterprises that anchor their NHI governance programs to a single large platform may inherit that vendor’s blind spots, pricing leverage, and product lifecycle decisions.
• Organizations should treat the current consolidation wave as a forcing function to document their NHI security requirements explicitly, evaluate coverage gaps before acquisition integrations are complete, and develop governance practices that do not depend on a single platform’s continued investment.

Background

Non-human identities are the credentials and access grants that enable software systems — rather than individual people — to authenticate and act within enterprise environments. The category spans a wide spectrum: long-lived API keys embedded in CI/CD pipelines, service accounts used by cloud workloads to access storage and databases, OAuth tokens delegating user access to third-party applications, certificates authenticating microservices within a service mesh, and, increasingly, the session tokens and tool-access grants that underpin autonomous AI agents. In practice, these identities are typically provisioned by developers or platform engineers during system build-out, rarely reviewed by identity governance teams, and seldom decommissioned when the systems they serve are deprecated or repurposed.

Astrix Security was founded in 2021 to address precisely this problem. The company built a platform specifically designed to discover and govern NHIs across enterprise environments — providing real-time inventory of API connections, OAuth grants, service accounts, MCP servers, and AI agents alongside risk scoring based on privilege scope, activity patterns, and policy violations. Astrix’s platform added lifecycle management capabilities — tracking credentials from provisioning to decommissioning — and threat detection oriented around the behavioral signatures of compromised NHIs, such as abnormal API call patterns or out-of-scope agent actions. After raising $45 million in a Series B round in late 2024, the company positioned itself as the definitive NHI security platform for enterprises accelerating AI agent deployment [5].

Cisco’s announcement of its intent to acquire Astrix on May 4, 2026 reflects a strategic decision to add NHI and AI agent governance to its security platform rather than build those capabilities organically. Cisco plans to integrate Astrix’s technology into Cisco Identity Intelligence, Cisco Secure Access, Duo Identity and Access Management, and its Splunk SIEM ecosystem, creating a unified view of both human and non-human identity activity across the enterprise [1]. The transaction was Cisco’s second AI security acquisition in recent weeks, following its earlier acquisition of Galileo Technologies, a provider of AI model observability tools [4].

Cisco’s announced intent to acquire Astrix came within months of CrowdStrike’s announcement, in January 2026, of its planned $740 million acquisition of SGNL [3][13]. SGNL’s approach to NHI and AI agent access is complementary but distinct: rather than discovery and lifecycle governance, SGNL focuses on runtime access enforcement, continuously evaluating whether identities — human, machine, or AI agent — should be granted access to a resource at the exact moment of request, based on live risk context rather than static role assignments. Together, the two acquisitions represent the absorption of two functionally different but thematically adjacent NHI security approaches into CrowdStrike’s Falcon platform and Cisco’s security stack, respectively.

The pace and scale of this consolidation are consistent with a market entering a rapid maturation phase. The global NHI access management market was valued at approximately $11.3 billion in 2025 and is projected to reach $12.2 billion in 2026, with a projected compound annual growth rate of approximately 12.2% through 2036 [7]. Purpose-built NHI vendors including Oasis Security, Entro Security, and Aembit continue to operate independently, but the acquisitions of Astrix and SGNL signal that major security platform vendors regard NHI as a strategic capability rather than a niche market — and are willing to pay meaningful acquisition premiums to acquire leading teams and technology quickly rather than waiting for organic development cycles. Astrix, for instance, was reported at a price range of $250 million to $350 million in pre-announcement speculation before the final $400 million figure was confirmed [14], underscoring the strategic value Cisco placed on the NHI capability.

Security Analysis

The NHI Threat Landscape

The urgency behind NHI security investment is grounded in an increasingly well-documented threat reality. Astrix co-founders stated that agents and non-human identities outnumber humans by roughly 100 to 1 within enterprise environments, and that they remain largely below the radar of security operations — representing what they characterized as “the biggest blind spot in our identity perimeter” [4]. The disproportion matters because most enterprise identity governance programs were designed around human login-and-logout patterns: access reviews are conducted for people, behavioral baselines are built around human activity, and conditional access policies are triggered by human authentication signals such as device health or geographic anomaly. NHI activity operates outside these patterns by design and largely passes through human-centric detection unexamined.

The attack surface created by unmanaged NHIs is not theoretical. OAuth tokens issued to AI SaaS tools during employee self-service signups commonly persist indefinitely, frequently survive employee departures, and often carry the full permission scope of the authorizing user — creating a class of dormant but active credentials that attackers can redeploy months after the original authorization intent has lapsed. Service accounts provisioned during application development often accumulate permissions over time as workload requirements change, arriving at privileged states that no single engineer intentionally created. AI agents deployed for workflow automation are frequently granted access to corporate systems during setup and, in many deployments, operate without any ongoing monitoring of whether their actions remain within intended scope. Each of these patterns represents an NHI category that traditional identity governance tools were not designed to inventory, evaluate, or govern.

CyberArk — a privileged access management vendor with a direct commercial interest in positioning AI agents as requiring enterprise PAM solutions — noted in its December 2025 threat analysis that as AI models evolve from simple assistants into complex multi-agent systems capable of executing actions across enterprise infrastructure, every agent becomes a privileged identity requiring the same lifecycle governance as a human privileged account — with the additional complication that agents can act at machine speed, amplifying the blast radius of a compromised credential far beyond what a human attacker operating manually could achieve [8].

What Astrix Brought to Market

Astrix’s platform addressed four functional capabilities that enterprise security teams have historically had to assemble from multiple point solutions or build in-house. First, its discovery engine maintained a real-time inventory of every NHI across connected SaaS, cloud, and on-premises environments — including API integrations, OAuth grants, service accounts, and AI agents — with contextual risk scoring based on privilege scope and operational activity. Second, its governance layer provided lifecycle management from provisioning through decommissioning, enabling security teams to identify over-privileged NHIs, enforce least-privilege principles, and automatically flag credentials approaching revocation criteria. Third, its threat detection capabilities generated behavioral baselines for NHI activity and alerted on deviations consistent with compromise or misuse — including out-of-scope agent actions and anomalous API call patterns. Fourth, Astrix offered centralized secrets management integration, aggregating visibility across credential vaults and surfacing credentials stored outside sanctioned vault infrastructure.

The coherence of these capabilities within a single purpose-built platform was Astrix’s primary market differentiator. Many enterprises had partial visibility into NHIs through existing identity governance tools, cloud-native IAM services, or CSPM platforms, but assembling that visibility into a unified risk picture required manual effort that security teams rarely had capacity for. Astrix made NHI governance operable at enterprise scale for teams without dedicated NHI engineering resources.

Platform Absorption and the Concentration Risk Question

The planned integration of Astrix’s capabilities into Cisco’s security platform is designed to create real benefits for organizations already operating within the Cisco ecosystem. Unified visibility across human identity (Duo IAM), NHI activity (Astrix), and network and application behavior (Cisco Secure Access) is intended to reduce the analyst effort required to correlate an identity event with downstream impact. Integration with Splunk would create a path for NHI threat signals to join the same detection logic applied to endpoint and network data, rather than being siloed in a separate console. For enterprises that have made significant investments in the Cisco security stack, the Astrix acquisition likely represents net capability gain without additional vendor relationships to manage.

However, the consolidation dynamic introduces structural risks that security programs should evaluate explicitly rather than absorb passively. When a purpose-built NHI security vendor is acquired by a large platform company, the acquired product’s roadmap typically competes for engineering and product resources against the acquiring vendor’s full portfolio. Integration timelines often extend; niche capabilities that do not align with the platform’s strategic priorities may be deprioritized; and the pricing model frequently shifts from the startup’s usage-based or seat-based pricing toward the platform’s enterprise licensing structures, which may increase cost or limit flexibility for teams that need NHI coverage without a full Cisco commitment.

More broadly, enterprises that build their NHI governance programs around a single platform’s implementation of discovery, governance, and threat detection are implicitly accepting that platform’s coverage decisions as their own. No platform comprehensively covers every NHI type, every cloud environment, or every AI agent framework. A security program that outsources its NHI threat model to a vendor’s product roadmap loses the organizational capacity to identify gaps independently — a capacity that becomes critical when a new agent framework or credential type emerges faster than platform updates.

This is the structural analogue to what security architects describe as vendor concentration risk in cloud infrastructure: the efficiency gains of consolidation are real, but they come at the cost of the independent evaluation capacity that would identify when the platform’s coverage no longer matches the organization’s threat surface.

Remaining Independent Vendors and Evaluation Urgency

Organizations evaluating their NHI security posture should recognize that the current window — before the Cisco–Astrix and CrowdStrike–SGNL integrations are complete — is a useful moment for independent assessment. Astrix and SGNL both continue to operate as standalone products during their respective integration periods, meaning their capabilities can still be evaluated on their own merits rather than as components of a larger platform. Purpose-built NHI vendors that remain independent — including Oasis Security, Entro Security, and Aembit — continue to offer alternatives for organizations seeking best-of-breed coverage or platform-neutral NHI governance.

The risk calculus shifts after integration is complete. Once Astrix’s discovery engine is tightly coupled to Cisco Identity Intelligence, evaluating it independently from the broader Cisco platform may not be practically feasible. Organizations that make NHI governance decisions based on post-integration capability bundles are making procurement decisions simultaneously with architectural ones — a pattern that historically constrains security program flexibility.

Recommendations

Immediate Actions

Security teams should conduct an NHI inventory now, before the Cisco–Astrix integration changes the available tooling landscape. An initial inventory does not require a commercial NHI platform: Google Workspace and Microsoft 365 administrators can enumerate OAuth grants through native admin consoles; cloud IAM services provide service account inventories through their respective APIs; and CI/CD pipeline administrators can identify embedded API keys through secrets scanning tools. This inventory will surface the most over-privileged NHIs and provide a baseline against which the completeness of any future platform-based discovery can be measured.

Organizations currently using Astrix Security as a standalone product should open conversations with their Cisco account team about integration timelines, licensing implications, and the roadmap for capabilities they depend on. Vendor acquisitions routinely include transition periods during which the acquired product continues to operate under its existing commercial terms, but those periods are finite, and understanding the timeline enables proactive planning.

Short-Term Mitigations

Security programs should document their NHI security requirements independently of any specific vendor’s current capabilities. This means articulating what types of NHIs exist in the environment (or are expected to exist as AI agent deployment scales), what discovery and governance capabilities are required to manage them, and what threat detection behaviors are necessary to identify compromise or misuse. That requirements document becomes the evaluation criterion against which both consolidated platform capabilities and independent vendor alternatives can be assessed objectively — rather than allowing vendor roadmaps to define the requirements retroactively.

Identity governance programs should be extended to explicitly include NHI lifecycle reviews alongside human access reviews. Adding a quarterly or semi-annual NHI review to existing access certification processes — reviewing service accounts, OAuth grants, and agent credentials for continued necessity, appropriate privilege scope, and active use — addresses the most common accumulation risk: dormant credentials and over-permissioned service accounts. It does not replace continuous behavioral monitoring for active NHI compromise, but it closes a meaningful gap in governance coverage without requiring dedicated NHI platform investment. Many identity governance platforms can scope access reviews to include non-human account types with modest configuration changes.

Strategic Considerations

The consolidation of NHI security into major platform vendors is likely to continue. Market analysts formally recognized machine identities as a distinct market segment by 2025 [7], and the acquisition activity concentrated in late 2025 and early 2026 reflects both that market validation and the platform vendors’ competitive response to it. Security programs that wait for the market to stabilize before addressing NHI governance will likely find themselves inheriting whatever coverage decisions the dominant platforms made — with limited ability to influence the outcome.

The more durable approach is to treat NHI governance as an organizational discipline rather than a product feature. Platform tools provide scale and efficiency, but the underlying governance principles — least privilege for all identity types, lifecycle management from provisioning through decommissioning, behavioral monitoring for anomaly detection, and inclusion of NHIs in access review processes — are vendor-neutral. Organizations that build those practices before committing to a platform are better positioned to evaluate whether a platform’s coverage is adequate, identify gaps the platform does not address, and retain the governance capacity to adapt as the NHI threat landscape continues to evolve.

As AI agent deployments scale across enterprises in 2026, the NHI problem is likely to grow faster than any single platform’s roadmap can fully address. Cisco’s acquisition of Astrix reflects the market’s recognition of that growth trajectory. Enterprise security programs should match that recognition with governance investment that is robust enough to survive the next round of acquisitions as well.

CSA Resource Alignment

The NHI security challenges addressed by the Cisco–Astrix acquisition are directly mapped to several CSA frameworks and publications that provide implementation guidance independent of any specific vendor.

MAESTRO (Multi-Agent Environment, Security, Threat, Risk, and Outcome) provides a seven-layer threat modeling framework for agentic AI deployments [9]. Layers 4 (Deployment and Infrastructure) and 7 (Agent Ecosystem) are particularly relevant to NHI governance: Layer 4 addresses the runtime environments in which agent credentials and secrets are stored and accessed, while Layer 7 addresses the trust relationships between agents and the external systems and APIs they interact with. Organizations applying MAESTRO to their AI deployments should incorporate NHI inventory and lifecycle governance as explicit requirements within both layers, regardless of which platform they use to implement that governance.

Agentic AI Identity and Access Management: A New Approach [10], published by CSA in August 2025, introduces a purpose-built IAM framework designed for the autonomy, ephemerality, and delegation patterns of AI agents. The document addresses the limitations of OAuth 2.1, SAML, and OIDC when applied to multi-agent systems and provides guidance on implementing Decentralized Identifiers (DIDs) and Verifiable Credentials (VCs) for agent authentication — an architectural approach that is platform-neutral by design and reduces dependency on any single vendor’s identity implementation.

The AI Controls Matrix (AICM) [11] provides a comprehensive framework for implementing trustworthy AI across organizational, technical, and societal boundaries. The AICM’s identity and access management control domains — covering both human and non-human identity governance — give security programs a structured, auditable basis for NHI policy development that does not require specific platform commitments. Organizations using the AICM as a governance reference can evaluate any NHI platform against the AICM control requirements and identify coverage gaps objectively.

CSA STAR for AI [12] provides an assurance framework for AI system trust that includes identity and access management as a core assessment domain. Enterprises deploying AI agents can use STAR for AI to assess and communicate the maturity of their NHI governance practices to customers, auditors, and regulators — framing that governance in terms of outcomes and controls rather than specific vendor products.

Finally, CSA’s Zero Trust guidance is foundational to the architectural posture that addresses NHI risk at its root. The consolidation of NHI security into large platform vendors does not change the underlying principle: no identity, human or non-human, should carry persistent implicit trust. Every NHI should be subject to continuous verification of continued need, minimal privilege scope, and active use — the same principles that inform human identity governance under a Zero Trust model, applied consistently regardless of which platform implements the verification.

References

[1] Cisco. “Securing the Agentic Workforce: Cisco Announces Intent to Acquire Astrix Security.” Cisco Blogs, May 4, 2026.

[2] Calcalist Tech. “Cisco acquires AI security startup Astrix for $400 million.” Calcalist Tech, May 2026.

[3] CrowdStrike. “CrowdStrike to Acquire SGNL to Transform Identity Security for the AI Era.” CrowdStrike Press Release, January 8, 2026.

[4] Network World. “Cisco grabs Astrix to secure AI agents.” Network World, May 2026.

[5] SecurityWeek. “Cisco Moves to Acquire Astrix Security to Tackle Non-Human Identity Risks.” SecurityWeek, May 2026.

[6] Cisco. “Cisco AI Readiness Index 2025: Realizing the Value of AI.” Cisco, October 2025.

[7] Meticulous Research. “Non-Human Identity (NHI) Access Management Market — Global Market Size, Share and Trends Analysis Report.” Meticulous Research, 2026.

[8] CyberArk. “AI Agents and Identity Risks: How Security Will Shift in 2026.” CyberArk Blog, December 4, 2025.

[9] Cloud Security Alliance. “Agentic AI Threat Modeling Framework: MAESTRO.” CSA Blog, February 6, 2025.

[10] Cloud Security Alliance. “Agentic AI Identity and Access Management: A New Approach.” CSA Research, August 18, 2025.

[11] Cloud Security Alliance. “AI Controls Matrix.” CSA Research, July 10, 2025.

[12] Cloud Security Alliance. “CSA STAR for AI.” Cloud Security Alliance, 2025.

[13] CNBC. “CrowdStrike buys identity security startup SGNL for $740 million in latest deal push.” CNBC, January 8, 2026.

[14] SiliconAngle. “Report: Cisco could acquire AI agent security startup Astrix Security for $250M+.” SiliconAngle, April 10, 2026.