The five-layer trust stack for agent-to-agent coordination : a compositional architecture for verifiable, auditable, and deception-resistant autonomous agent interactions

Abstract

As autonomous artificial intelligence (AI) agents are deployed across enterprise and defense environments, ad hoc agent-to-agent coordination introduces machine-speed accountability gaps that existing access control and audit frameworks were not originally designed to address. Without a structured trust framework, agent interactions are susceptible to the same deception techniques that adversaries apply against human decision-makers: impersonation, authority spoofing, false commitments, and the manipulation of information presented as authoritative. The Five-Layer Trust Stack provides a composable, cryptographically grounded architecture that addresses these coordination risks in sequence: from establishing verifiable identity through scoped delegation of authority, machine-checkable policy enforcement, signed commitments, and tamper-evident audit logs with arbitration hooks. A foundational governance layer—Layer 0—anchors the entire stack in human authority, policy authorship, and real-time intervention capability. The stack does not replace existing Department of Defense ICAM (Identity, Credential, and Access Management), Zero Trust, RMF (Risk Management Framework), or audit systems. Rather, it composes these existing controls into a coordination-layer architecture suitable for autonomous, machine-speed interactions, including air-gapped and disconnected, intermittent, and limited (DIL) bandwidth environments.