Two Competing Theories of Agent Identity

As discussions around agent identity take center stage across the technology sector, two distinct structural philosophies have emerged. The FIDO Alliance concluded its Authenticate event by introducing a wallet-certification announcement, framing identity as a characteristic carried by a human user. Meanwhile, AGNTCon and MCPCon opened in San Jose with an agenda positioning identity as a property issued directly by a platform.

FIDO's approach aims for an ecosystem where users and their agents interact with services via a verifiable human identity. According to FIDO, this cryptographic root of trust serves as a prerequisite for secure agentic commerce, ensuring every transaction is linked back to verified human authorization. The alliance intends to publish a Credential Manager Reference Architecture and a Framework for Verifier Identification in the coming weeks.

Conversely, the platform-centric model discussed in San Jose focuses heavily on the tool-call layer. Presenters outline patterns involving the interception of agent calls and the substitution of short-lived scoped credentials instead of permanent secrets. While FIDO moves identity outward from the individual, the platform model issues identity downward from the system.

The Underlying Verification Gap

Despite the distinct approaches offered by both models, a production vulnerability recently exposed a blind spot shared by both theories. The BeyondTrust Remote Support breach demonstrated an improper origin validation flaw where an internal messaging component trusted message origins rather than verifying them. Although downstream identity layers functioned as intended, the lack of request origin verification allowed an unauthenticated attacker to execute file operations.

Even FIDO's proposed Verifier Identification framework acknowledges limitations, conceding that credential managers excel at verifying user identity but lack scalable mechanisms to validate the origin of requests from verifiers. Industry observers note this ongoing challenge remains a primary hurdle for enterprise adoption as Q4 budgets face increasing scrutiny.

Shipped Perimeters and Enterprise Adoption

While overarching interoperability standards remain in draft form, various organizations have rolled out proprietary governance stacks. NVIDIA introduced the OpenShell Reference Architecture, embedding enforcement directly into silicon using hardware components like Vera CPUs and BlueField-4 DPUs. Similarly, health-native gatekeepers and secure invoice payment pilots involving Visa, BILL, and Billtrust have gone live to handle structured B2B financial workflows.

These developments illustrate a fragmented landscape where organizations must choose between adopting existing proprietary perimeters or waiting for broader, standardized interoperability frameworks to reach general availability.