TrustedARI: Towards Trust-Native Agentic Routing Infrastructure for Agentic AI

TrustedARI introduces a novel trust-native agentic routing infrastructure designed to mitigate trust risks in AI agents accessing external models and services. It features an ARI-adapted three-party TLS handshake that reduces communication overhead by 39.34%, a privacy-preserving query-construction protocol with minimal computational and communication costs (0.19 seconds and 0.58 MB), and a verifiable billing protocol that accelerates proof generation by 28.20x. This architecture enhances security and efficiency for practitioners developing agentic AI systems without requiring changes to existing service providers.