On July 9, 2026, the International Telecommunication Union announced a new effort to develop international frameworks for the identity and trustworthiness of autonomous AI agents, forming a Focus Group on Trust and Identity for Humans and Agentic AI. It's the first UN-agency-level attempt to standardize how software agents prove who they are across borders and platforms, rather than leaving that to individual vendors. The stated scope is broad: the group is expected to study common terminology, identity and trust architectures, agent discovery, credential interoperability, lifecycle models, security criteria, benchmarks, and a roadmap for future standards. That list matters because it treats identity, discovery, and trust as one connected problem rather than three separate ones — an agent's identifier is only useful if other systems can discover it, verify who issued it, and check whether it's still valid. The real fight isn't over whether agents need credentials — nobody disputes that anymore. It's over who controls issuance and recognition. A workable system needs to support multiple issuers, pseudonymous identities, selective disclosure, short-lived permissions, transparent revocation, and offline validation. Get that wrong and you don't get an open standard — you get a gatekeeper. An open technical spec can still produce a closed ecosystem if the handful of platforms and institutions that matter only accept credentials from a short list of approved issuers. That's the practical risk with any agent-passport scheme: the spec can be neutral while the market for who gets to sign it isn't. This lands amid a busy few months for adjacent efforts — NIST's Center for AI Standards and Innovation opened its own concept paper on agent identity and authorization for comment earlier this year, Microsoft shipped Entra Agent ID as a general-availability product, and a DNS-anchored identity proposal from Identity Digital went to the IETF in June. The ITU's move doesn't replace any of that; it's an attempt to keep the international layer from fragmenting the same way regional AI regulation already has. Whether it succeeds will depend less on the technical architecture — most of the building blocks (OAuth extensions, verifiable credentials, DNS anchoring) already exist — and more on whether governments and large platforms agree to recognize credentials issued outside their own walled gardens. For agent developers, the immediate takeaway is narrower and more practical: stop reusing human credentials for agents. Give each production agent scoped, short-lived authority tied to a specific task and resource, because whatever international framework eventually lands, it will be built on top of that discipline, not instead of it.