Non-Human Identity Management Emerges as Critical Governance Challenge
The News
At KubeCon North America 2025, AuthZed discussed its enterprise authorization platform built on SpiceDB, an open-source Zanzibar-equivalent system with approximately 6,000 GitHub stars that commercializes Google’s Zanzibar model through dedicated deployments for enterprises like banks and cloud self-service offerings.
The company serves major customers, including OpenAI’s ChatGPT Enterprise connectors at scale, handling tens of billions of fine-grained permissions, replacing brittle homegrown authorization stacks with performance, low latency, and broad regional coverage as key differentiators versus competitors like OpenFGA and bundled authorization from Okta/Auth0. AuthZed targets two primary customer segments: large enterprises with complex, high-scale authorization needs and growth-stage vendors pursuing enterprise readiness who need to meet buyers’ bespoke authorization requirements, such as unusual role constructs to unlock deals.
Analyst Take
AuthZed emphasizes treating AI agents as first-class non-human identities within its authorization model, enabling consistent authorization, traceability, and accountability for agent workflows alongside human users. The company positions this capability as addressing enterprise AI viability challenges where many projects fail partly due to security and authorization gaps, particularly concerns around MCP servers and agent-to-agent data extraction, breaching governance of heritage systems, and undermining security and enterprise confidence.
While AuthZed does not provide identity management itself, the platform allows customers to define “user” types where agents can be modeled as non-human identities, supporting controls like human-in-the-loop approvals and agent action attribution. The typical adoption pattern is “almost never greenfield,” with deployments starting by solving a new, hard use case before expanding across products—for example, Turo’s co-hosting and fleet permissions requiring intricate graph-like relationships that existing systems could not handle, with AuthZed solving the initial use case before scaling wider.
AuthZed is a series A-stage company, backed by General Catalyst, Work-Bench, Y Combinator, and Amplify, and is preparing for its next funding round. The company is developing channel partnerships, with CEO Jake describing AuthZed as a “database-like” component that customers integrate without branding, citing IBM organically introducing AuthZed to the National Bank of Canada as “the authorization solution” as validation of partner-led delivery potential.
AuthZed’s positioning of SpiceDB as enterprise-grade Zanzibar-style authorization addresses a genuine market gap where organizations recognize that embedded RBAC and homegrown permission systems do not scale to complex, fine-grained authorization requirements, but they lack the expertise and resources to build Google-class authorization infrastructure internally.
The Zanzibar model, using relationship tuples and graph-based permission checks, provides flexibility for modeling complex organizational hierarchies, resource ownership, and delegated permissions that traditional role-based systems cannot express. However, the abstraction level and conceptual complexity of relationship-based authorization create adoption barriers for developers accustomed to simpler RBAC patterns, requiring AuthZed to balance the power and flexibility that attracts enterprise customers against the learning curve and operational complexity that slows initial adoption.
The emphasis on performance, low latency, and regional coverage as differentiators reflects operational requirements for authorization systems that must answer permission checks on every API request without becoming bottlenecks. Authorization latency directly impacts application response times, making sub-10ms P99 latency critical for user-facing applications, while regional deployment addresses data sovereignty requirements where organizations cannot send authorization data across borders.
However, these operational characteristics are table stakes rather than sustainable differentiation; competitors will match performance and coverage over time, forcing AuthZed to demonstrate ongoing value through superior developer experience, ecosystem integrations, or unique capabilities like the AI agent security positioning. The comparison to OpenFGA (also Zanzibar-inspired and open source) and bundled authorization from identity providers creates a three-way competitive dynamic where AuthZed must prove that dedicated authorization infrastructure justifies the operational complexity versus embedded solutions or alternative open-source implementations.
The AI agent security positioning, treating agents as first-class non-human identities, addresses an emerging governance challenge but requires validation that organizations recognize this as a priority problem worth solving with dedicated infrastructure. The concern about MCP servers and agent-to-agent data extraction breaching governance of heritage systems reflects legitimate security risks, but the market maturity of agentic workflows remains uncertain, with many organizations still experimenting with AI agents rather than deploying them at scale in production environments with sensitive data.
AuthZed’s capability to model agents as non-human identities with consistent authorization, traceability, and accountability provides a technical foundation for agent governance, but success depends on whether enterprises adopt this architectural pattern versus treating agents as extensions of human users or implementing ad-hoc controls. The human-in-the-loop approval and agent action attribution capabilities address compliance and audit requirements, but organizations must determine whether fine-grained authorization for agents provides sufficient governance or whether additional controls around agent behavior, data access, and decision-making are necessary.
The adoption strategy, starting with novel, hard use cases before expanding across products, reflects a pragmatic go-to-market that addresses the greenfield deployment challenge; organizations rarely replace working authorization systems wholesale due to risk and effort. The Turo example, co-hosting and fleet permissions requiring graph-like relationships, demonstrates the type of complex authorization scenario where SpiceDB’s relationship model provides clear value over simpler RBAC systems.
However, this adoption pattern also creates scaling challenges as AuthZed must repeatedly prove value for each new use case rather than achieving broad platform adoption, and the incremental expansion depends on initial use case success, creating organizational champions who advocate for wider deployment. The targeting of growth-stage vendors needing enterprise readiness addresses a different market segment where authorization complexity becomes a sales blocker, but these customers may have different requirements around multi-tenancy, customization, and operational simplicity than large enterprises with dedicated platform teams.
Looking Ahead
AuthZed’s success depends on whether the next 12-18 months validate that fine-grained, relationship-based authorization becomes a recognized enterprise requirement rather than remaining a specialized solution for organizations with unusually complex permission models. The company must demonstrate that the operational complexity and learning curve of Zanzibar-style authorization deliver sufficient value, through better security, more flexible permission models, or reduced development effort, to justify adoption over simpler alternatives. The AI agent security positioning provides a differentiation opportunity, but it requires that agentic workflows achieve production maturity with governance requirements that existing authorization systems cannot meet, rather than remaining experimental deployments where ad-hoc controls suffice.
The competitive landscape for authorization is evolving as identity providers expand beyond authentication into authorization, cloud platforms build native permission systems, and alternative open-source Zanzibar implementations mature. AuthZed’s differentiation through SpiceDB’s open-source foundation aligns with research indicating 68% of organizations prefer vendors backing open source with full support, but success requires building ecosystem momentum where SpiceDB becomes the default choice for relationship-based authorization rather than one option among several.
The channel partnership strategy, positioning AuthZed as a “database-like” component integrated without branding, addresses the skills gap and complexity barriers, but it also creates dependency on partners to drive adoption and deliver implementation services. AuthZed must balance direct enterprise sales that build brand recognition and capture high-value deals against partner-led distribution that scales reach but reduces control and margins, while proving that authorization infrastructure justifies dedicated investment rather than remaining embedded within broader identity, security, or application platforms.
High Availability Becomes a Strategic Control Layer in 2026 IT Architectures
NKP 2.17 Signals a Security-First Turn in Enterprise Kubernetes Platforms
Omni Agent Intelligence Signals the Next Phase of AI-Native CX Governance
Always-On Retail Becomes a Platform Requirement at NRF 2026
Agentic Commerce Moves From Concept to Control Plane at NRF 2026
