CrowdStrike and IETF Move in Parallel:
AI Agent Identity Is Becoming Its Own Security Category

In July 2026, two independent events point in the same direction: AI agent identity management is becoming a standalone security category. CrowdStrike launched a dedicated "Continuous Identity for AI Agents" product, while the IETF published the Agent Identity Protocol (AIP) draft, defining a decentralized identity standard for agents.

CrowdStrike: AI Agents Need Continuous Identity Verification

CrowdStrike's "Continuous Identity for AI Agents" is built on a simple premise: traditional human user identity models do not fit AI agents. Agents are non-human, automated, short-lived entities with behavior patterns completely different from humans — an agent may execute hundreds of operations in seconds, jumping from one context to another.

The solution provides continuous identity verification for every AI agent, rather than static one-time identity assignment. Key features include:

• Continuous verification. Agent identity is not static — re-verify at every operation
• Behavioral baselines. The system learns each agent's normal behavior pattern and alerts on deviation
• Real-time revocation. Revoke agent access immediately upon anomaly detection, without waiting for credential rotation
• Cross-platform compatibility. Supports multiple agent frameworks and runtime environments

CrowdStrike's entry is a significant signal. When one of the world's largest cybersecurity companies treats AI agent identity as an independent category, enterprise security teams should pay attention.

IETF AIP: Building a Decentralized Identity Standard for Agents

Almost simultaneously, the IETF published the Agent Identity Protocol (AIP) draft (draft-singla-agent-identity-protocol-03), proposed by P. Singla. AIP defines a decentralized identity, delegation, and authorization framework for AI agents, combining W3C Decentralized Identifiers (DIDs), capability-based authorization, cryptographic delegation chains, and deterministic validation.

AIP's core design principles include:

• Decentralization. No central identity authority — agents can autonomously generate and verify identities
• Capability-based authorization. Agents carry "capability tokens" proving access to specific resources, rather than sharing credentials
• Delegation chains. One agent can delegate capabilities to another, forming auditable delegation chains
• Deterministic verification. All identity verification steps are deterministic and machine-executable without human intervention

This is especially critical for multi-agent collaboration. When multiple agents need to work together to complete a task, AIP ensures each agent can verify the identity and permission scope of others without sharing credentials or relying on centralized identity systems.

Why Agent Identity Management Differs from Traditional IAM

Traditional Identity and Access Management (IAM) systems were designed for human users. Humans have stable identities, fixed working hours, and predictable behavior patterns. AI agents are fundamentally different:

• Non-human. Agents don't "log in" — they execute operations via APIs and tool calls
• Automated. Agents make decisions and execute actions in milliseconds
• Short-lived. Many agents are created for specific tasks and destroyed upon completion
• Cascading. One agent may call another, forming complex invocation chains

These differences mean you cannot simply apply human IAM policies to AI agents. A dedicated identity model for agents is needed — exactly what CrowdStrike and the IETF are building.

Implications for Enterprise AI Governance

Both events point to a clear direction: AI agent identity management is becoming the infrastructure layer of enterprise AI governance. Just as the internet era needed IP addresses and DNS, the AI era needs agent identity protocols and continuous verification systems.

For enterprises deploying AI agents, now is the time to start building agent identity management. You don't need to wait for standards to fully mature — CrowdStrike's commercial solution and the IETF draft already provide a clear roadmap. Start with one independent identity per agent, then gradually build continuous verification and capability-based authorization. This will lay a solid foundation for enterprise AI governance.