AgentBound: Verifiable Behavioral Governance for Autonomous AI Agents
06:00 · July 1, 2026 · arXiv cs.AI RSS
Autonomous AI agents increasingly perform consequential actions on behalf of human principals, including financial transactions, external communications, and enterprise workflows. Existing agent infrastructure relies on identity federation and delegated authorization to authenticate workloads and control resource access, but it cannot determine whether an authorized action should be executed under the current behavioral and operational context. We present AgentBound, a runtime governance framework that provides verifiable behavioral oversight for autonomous AI agents. AgentBound evaluates each proposed action using three independent authorities: delegated authorization, owner-signed behavioral constitutions, and site action contracts. Their judgments are conservatively composed through a formal decision model to determine whether an action should be permitted, reviewed, or denied before execution. To provide accountability, AgentBound generates cryptographically verifiable governance receipts that bind every action to the exact delegation, policy, and semantic artifacts governing the decision, enabling independent replay verification and policy provenance. The framework also introduces standing delegation for long-running agents, allowing periodic workloads to operate under continuously refreshed governance policies while preserving revocability and bounded authority. We present the formal foundation, system architecture, governance receipt protocol, and AgentBound-Bench, a benchmark framework for evaluating governance correctness, authority composition, and accountability. Rather than replacing model alignment, AgentBound complements it by providing a deterministic governance layer between authorization and execution, transforming governance from a process that must be trusted into one that can be independently verified.
Summary
AgentBound addresses a persistent gap in autonomous AI agent systems, where identity federation and delegated authorization can authenticate workloads and restrict resource access yet leave behavioral suitability unexamined. The framework inserts a non-bypassable runtime governance layer that evaluates every proposed action against three independent authorities: existing delegated authorization, owner-signed behavioral constitutions that encode principal intent, and site-specific action contracts that capture contextual constraints. These judgments are combined through a conservative formal decision model that returns a deterministic verdict of permit, review, or deny before execution proceeds.
To support accountability, AgentBound produces cryptographically signed governance receipts that bind each action to the precise delegation, policy, and semantic artifacts active at decision time. The receipts enable independent third-party replay verification and policy provenance without reliance on the agent’s internal state. The design also introduces standing delegation, allowing long-running or periodic agents to operate under continuously refreshed policies while maintaining revocability and bounded authority.
The paper supplies the underlying formal model, system architecture, receipt protocol, and AgentBound-Bench, a benchmark suite for assessing governance correctness, authority composition, and auditability. Rather than supplanting model alignment techniques, AgentBound supplies a complementary deterministic enforcement mechanism that converts governance from an implicit trust assumption into a verifiable process.
Why it matters
Directly addresses verifiable behavioral governance and accountability for AI agents, aligning with Dutch/EU priorities on ethical, transparent AI and upcoming regulation. Provides formal models, receipts, and a benchmark that Dutch researchers and advanced practitioners can evaluate or extend for enterprise and regulatory use cases.
