The study develops a three-layer framework of agent-readability, traceability, and governability, theorizes agent-mediated contributions as governable boundary objects, and advances compliance-enabling digital innovation governance while preserving maintainer decision authority.
Abstract
Generative AI and coding agents are intensifying a central governance tension in open-source software (OSS): they scale contribution generation faster than maintainers can assess risk, evidence, and accountability. Existing responses improve agent-readability and traceability, but project rules must also organize contribution-specific risk, evidence, accountability, and review-gate states. We theorize this organizational arrangement as project-side governability infrastructure. A diagnostic audit of 50 GitHub repositories finds widespread general governance artifacts, observable agent-readability, and fragmented AI-governance cues, but no project-wide arrangement that coordinates shared rules, preparation obligations, verification rights, and maintainer decision authority across AI-mediated contribution workflows. We develop the Agent Governance Manifest (AGM) as a repository-hosted boundary resource and bidirectional governance contract linking contributor-side evidence preparation with maintainer-side verification. In a controlled reviewer-side evaluation with 15 participants and 75 task-level outputs, AGM-supported materials improved exact risk-label recovery (37/38 vs. 15/37) and perceived review support (6.14 vs. 3.27 on a 1-7 scale). In a contributor-side feasibility check, 15 participants completed 45 tasks; all final packages represented the core governance state correctly, and 41 passed strict structural validation. The study develops a three-layer framework of agent-readability, traceability, and governability, theorizes agent-mediated contributions as governable boundary objects, and advances compliance-enabling digital innovation governance while preserving maintainer decision authority.
Agentic AI systems, LLM-based agents that autonomously plan multi-step tasks and execute them via tool calls, are increasingly deployed in organizations, yet systematic instruments to operationalize governance, transparency, and accountability for agentic tool usage remain lacking. This paper presents the conceptual design of a trace-based governance framework that operates without access to internal model reasoning (Chain-of-Thought) and instead leverages observable execution evidence. The core artifact is an MCP (Model Context Protocol) Governance Gateway, a platform-agnostic integration layer positioned between AI host applications and organizational tools. The framework addresses three governance dimensions: (1) transparency as observability through structured evidence capture of tool calls, data provenance, and human approvals; (2) human oversight through risk-based policy mechanisms including approval gating and capability boundaries; and (3) accountability through auditable logs with tamper-evident integrity mechanisms. We formalize governance requirements derived from the EU AI Act and EU HLEG Trustworthy AI guidelines, propose an Agent Action & Evidence Graph as a formal evidence model, and outline a three-stage evaluation strategy. The framework contributes a novel approach to operationalizing regulatory AI governance requirements for heterogeneous agentic tool ecosystems.
Heidrun Mühle, Andrea Stöckl· 2026 6th International Confe...· 0 citations
Large language models are increasingly used to review, clarify, rewrite, and trace software requirements. These applications create a governance problem that output-quality assessment alone cannot resolve: a fluent proposal may rely on inadmissible evidence, alter stakeholder intent, introduce unsupported specificity, or imply an organizational commitment that the model has no authority to make. Existing work on retrieval-augmented generation, controlled natural language, formal verification, human oversight, and AI governance supplies relevant controls, but it does not specify the procedural status of an individual LLM proposal relative to a controlled requirements artifact. This article develops an artifact-centered, human-in-the-loop framework in which the permitted influence of a proposal is the primary object of governance. The framework combines five governance functions—governed evidence, bounded context construction, controlled LLM analysis, pre-commit verification, and accountable human approval—with four artifact-influence states: A0 advisory observation, A1 evidence-linked candidate, A2 verified recommendation, and A3 approved and committed change. Its central theoretical claim is that output quality, evidential legitimacy, verification status, and authority to commit a change are distinct properties and should not be collapsed into a single confidence judgment. Seven falsifiable hypotheses translate the model into measurable comparisons involving source admissibility, context leakage, unsupported specificity, semantic drift, reviewer agreement, unreviewed changes, governance cost, and organizational maturity. Human review is treated as both a necessary decision boundary and a potential source of automation bias, anchoring, and fatigue. The framework is conceptual rather than empirically validated and provides a basis for controlled experiments, field studies, and longitudinal evaluation.
Chuanjin Zhu· Advances in Engineering Inno...· 0 citations
Artificial intelligence is moving from analytical support toward active participation in enterprise decisions, creating an organization-design problem: firms must decide which rights may be delegated to AI and how responsibility should follow the actors who can prevent, challenge, or remedy failure. Existing work explains automation, augmentation, delegation, human oversight, and responsible AI governance, but does not reveal how specific transfers of decision authority create responsibility gaps inside a focal enterprise decision. This conceptual paper develops a contingency governance framework through a transparent theory-synthesis procedure. A purposive corpus of 44 peer-reviewed studies, standards, and regulatory sources was assembled through anchor studies, targeted keyword searches, and citation chaining. First-order authority and responsibility terms were coded, compared, and abstracted until two successive search iterations produced no new categories. The resulting framework distinguishes seven decision rights—information access, recommendation, selection, approval, veto, execution, and escalation—and five responsibility domains—system design, decision process, outcome stewardship, oversight, and remediation. Its central mechanism is rights-control-responsibility alignment: delegating a right shifts effective control and evidence access, while governance fails when the responsible actor lacks the competence, authority, or information to intervene. Decision exposure and AI autonomy determine four governance archetypes, while AI reliability conditions the permissible scope of selection and execution rights. Eight empirically testable propositions specify mechanisms, moderators, competing explanations, and falsification conditions. Two worked applications show how the architecture produces more precise governance than a generic human-in-the-loop requirement. The paper contributes a decision-level theory of enterprise AI governance and provides managers with an auditable method for allocating rights, responsibilities, evidence, and lifecycle controls.
Fang Sun· ICCK Transactions on Systems...· 0 citations
Enterprises are rapidly moving beyond static, single-turn Large Language Model (LLM) deployments toward knowledge-grounded systems — architectures such as Retrieval-Augmented Generation (RAG) that condition outputs on enterprise document stores — and toward agentic AI systems that plan, invoke external tools, and execute multi-step actions with limited human supervision. While prior work in enterprise AI assurance has addressed hallucination and demographic bias as discrete model-quality problems, the shift to grounded and agentic architectures introduces a qualitatively different class of governance risk: retrieval-grounding failure, tool-call malformation, permission-scope violation, and the compounding of small per-step errors into materially harmful multi-step outcomes. This paper characterizes these risks as structural properties of agentic enterprise systems rather than incidental model defects, and proposes the Agentic and Knowledge-Grounded Assurance (AKGA) framework, a six-layer governance pipeline that integrates grounding-fidelity verification, tool-call correctness checking, action-safety scoring, and a composite Agentic Risk Index (ARI) with tiered, threshold-based escalation to human review. Using a simulated benchmark spanning healthcare care-coordination, financial-services operations automation, and insurance claims-processing agents, the framework is shown to raise mean grounding fidelity, tool-call correctness, and action-safety scores from the 0.54–0.63 range to above 0.88 post-governance, while step-wise verification is shown to substantially suppress the compounding of risk across sequential agentic task chains relative to an ungoverned baseline. The paper concludes that validation and governance of grounded and agentic AI must be treated as a first-class enterprise reliability engineering discipline — auditable, thresholddriven, and embeddedacross the inference lifecycle — rather than as an extension of conventional model evaluation.
Suresh Babu Narra· International Journal of Int...· 0 citations
AI agents increasingly enter practitioner workflows through delegated, multi-step tasks, such as data analysis, document review, coding, and summarization. Existing governance debates tend to emphasize provider-level technical governance, which steers general model behavior, and policy, which defines the boundaries of legitimate use. Both are necessary, but neither fully specifies how domain-specific norms should guide the intermediate choices agents make during task execution. This article develops runtime configuration as a meso-level, agent-facing governance mechanism for this operational gap. Runtime configuration refers to persistent, inspectable, and revisable instructions and supporting materials loaded at use time that specify decision authority, documentation and evidence-preservation duties, and conditions for human escalation. These artifacts bridge domain practice and agent execution. They translate situated normative commitments into agent-facing guidance while connecting that guidance to technical controls, work outputs, and human review. We illustrate the framework through a case study of investigative journalism, comparing three conditions: an unconfigured baseline and two configured conditions that guided agent runs on a public-records data task. Across the runs, the clearest differences associated with configuration concerned the conditions of delegation rather than substantive accuracy: The runs differed in escalation, provenance, workflow recoverability, and the visibility of consequential decisions. The aim of runtime configuration is not to replace model alignment, policy, expertise, or institutional accountability. Instead, it makes situated delegation more inspectable by translating normative domain commitments into operational guidance for agentic work.
Nick Hagar, Nick Diakopoulos· AI and Ethics· 0 citations
This study aims to examine how digital policy and regulatory governance should respond to vendor-mediated generative artificial intelligence (AI) in regulated financial services. It argues that the central problem concerns not only model assurance but also the evidentiary pipeline through which customer data, vendor processing, generated outputs and human review become auditable.
The article uses a conceptual and design-orientated documentary comparison of Singapore and Vietnam. It analyses AI governance, data protection, financial supervision and third-party risk instruments through four functional axes, derives operational indicators from the documentary corpus and examines their internal coherence through a structured illustrative case in financial services.
Singapore’s interoperability-orientated model and Vietnam’s dossier-based model of legal visibility provide different regulatory entry points. Both remain incomplete unless institutions preserve workflow-level evidence across procurement, configuration, deployment, output verification and supervisory review.
The framework links risk triggers to pipeline maps, vendor due diligence, transfer records, output-verification protocols and audit trails.
The article develops a conceptually grounded pipeline accountability framework that connects vendor obligations, data movement, generated outputs and human verification. It is operationally specified but remains a design proposition requiring empirical testing and refinement.
V. Hoang, Ngoc Mai Nguyen· Digital Policy Regulation an...· 0 citations