Scholarlink

<p>This study proposes a paradox-aware governance framework for the dual-use risks of agentic artificial intelligence. Anchored in paradox theory as the primary conceptual lens, the framework draws supplementary insights from institutional logics, dynamic capabilities, affordance theory, and panarchy to address context-specific implementation challenges. A critical-comparative analysis of historical dual-use regimes spanning nuclear, biological, chemical, cyber, and export controls yields five core governance paradoxes: Capability Control versus Innovation, Autonomous Action versus Accountability, Algorithmic Opacity versus Verification, Centralized Oversight versus Decentralized Development, and Proactive Safety versus Information Hazards. The study employs directed content analysis with transparent coding procedures, documenting both deductive category application and inductive emergence. Real-world cases including the GPT-4 deployment, Meta’s LLaMA model proliferation, and the EU AI Act’s tiered risk framework ground the propositions empirically. Each paradox is operationalized through paired mechanisms, boundary conditions, and governance levers such as Adaptive Capability Ledgers and Behavioral Compliance Sandboxes. The framework advances five falsifiable propositions specifying testable relationships between governance interventions and measurable outcomes. Equity is explicitly addressed through concrete instruments including compute-credit redistribution protocols, equity-weighted governance metrics, and capacity-building requirements for Global South stakeholders. Validation pathways encompassing quantitative indicators and qualitative stakeholder feedback ensure adaptive refinement. This integrated model transforms static regulations into living systems that co-evolve with AI’s inherent malleability, offering regulators, practitioners, and scholars actionable tools to sustain innovation while preempting irreversible harms.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Dynamic governance for agentic AI: synthesizing dual-use regimes through paradox theory

  • 期刊论文

  • 发表日期：

    2026年5月31日

查看全文
  • Ofem Ofem

摘要

This study proposes a paradox-aware governance framework for the dual-use risks of agentic artificial intelligence. Anchored in paradox theory as the primary conceptual lens, the framework draws supplementary insights from institutional logics, dynamic capabilities, affordance theory, and panarchy to address context-specific implementation challenges. A critical-comparative analysis of historical dual-use regimes spanning nuclear, biological, chemical, cyber, and export controls yields five core governance paradoxes: Capability Control versus Innovation, Autonomous Action versus Accountability, Algorithmic Opacity versus Verification, Centralized Oversight versus Decentralized Development, and Proactive Safety versus Information Hazards. The study employs directed content analysis with transparent coding procedures, documenting both deductive category application and inductive emergence. Real-world cases including the GPT-4 deployment, Meta’s LLaMA model proliferation, and the EU AI Act’s tiered risk framework ground the propositions empirically. Each paradox is operationalized through paired mechanisms, boundary conditions, and governance levers such as Adaptive Capability Ledgers and Behavioral Compliance Sandboxes. The framework advances five falsifiable propositions specifying testable relationships between governance interventions and measurable outcomes. Equity is explicitly addressed through concrete instruments including compute-credit redistribution protocols, equity-weighted governance metrics, and capacity-building requirements for Global South stakeholders. Validation pathways encompassing quantitative indicators and qualitative stakeholder feedback ensure adaptive refinement. This integrated model transforms static regulations into living systems that co-evolve with AI’s inherent malleability, offering regulators, practitioners, and scholars actionable tools to sustain innovation while preempting irreversible harms.