UCII Voice Authority: Identity for AI Agents

AI agents can increasingly call tools, access infrastructure, spend money, and take real-world actions. UCII Voice Authority addresses a fundamental problem: who actually decides what an AI agent is allowed to do? Built with AssemblyAI's real-time Voice Agent technology and UCII (Universal Cryptographic Identity Infrastructure), the project gives a voice agent its own cryptographic identity and independently enforceable delegated authority. The core principle is simple: the LLM is not the authority source. During the live demo, a human speaks naturally with the AssemblyAI-powered agent and asks whether it has authority to inspect or purchase compute. The agent performs live UCII authorization checks instead of deciding for itself. The demo proves the complete authority lifecycle. The agent verifies authorized capabilities, a human requests revocation of its purchase authority, and a fresh UCII check returns DENIED. Another independent check proves that its compute inspection authority remains AUTHORIZED. Revoking one permission therefore does not destroy the agent's identity or disable everything it can do. Authority is granular, externally enforced, and independently verifiable. AssemblyAI provides the real-time conversational layer, including speech interaction and structured tool calling. UCII provides the cryptographic identity and authority layer behind it. This separates identity, intelligence, and authority. Instead of trusting an AI agent to decide what it may do, an external system verifies and constrains its permissions. The same architecture can govern autonomous agents using payments, APIs, enterprise infrastructure, robots, devices, and services. Give an AI an identity. Delegate only the authority it needs. Verify before it acts. Revoke individual capabilities without destroying the identity.