Cloud Security Alliance published a new whitepaper, "Agentic AI Identity and Access Management: A New Approach," that treats agent discovery as part of the identity stack rather than a separate concern bolted on afterward. The paper introduces a layered framework built around Decentralized Identifiers and Verifiable Credentials, and it names an "Agent Name Service" as one of its core components alongside dynamic access control and a unified session management layer. That's the detail worth sitting with. Most identity proposals stop at authentication and authorization — proving who an agent is and what it's allowed to do. CSA's framework folds naming and lookup into the same architecture, which implies that in a multi-agent system, an agent can't be trusted at all until it can be found, verified, and revoked through one consistent mechanism, not three disconnected ones. The paper's scope is broad by design. It covers the full lifecycle from identity creation and attestation through runtime authorization, logging, monitoring, and incident response, and it compares centralized, decentralized, and federated deployment models rather than picking a winner. It also leans on CSA's existing MAESTRO threat-modeling framework to analyze security considerations layer by layer — mapping identity failures against the seven layers MAESTRO already defines for agent ecosystems, from foundation models up through agent-to-agent interaction. The practical stakes: enterprises adopting multi-agent systems currently have to stitch together separate tools for agent naming, credential issuance, session revocation, and audit logging, each built for a different generation of infrastructure. A framework that treats these as one coherent lifecycle — rather than four vendor integrations — is what a real production deployment actually needs before anyone can say an agent's identity is trustworthy end to end. What's still missing is adoption evidence. CSA whitepapers propose architecture; they don't mandate it, and past CSA frameworks like MAESTRO have taken months to see real implementation traction, mostly as an analytical lens applied retroactively to systems like OpenAI's Responses API and Google's A2A protocol rather than as a deployed standard. Whether the Agent Name Service or the unified session layer gets built into actual products, or stays a reference architecture cited in other papers, is the open question for the next few months. The honest read: this is a synthesis document, not a new primitive. Its value is in refusing to let discovery, credentialing, and revocation be treated as separate problems solved by separate teams.