Leveraging Human Rights Frameworks for Agentic AI Governance

Abstract

Agentic artificial intelligence (AI)—systems that reason, act, and interact with autonomy and persistence—presents qualitatively new governance challenges compared to traditional large language models. By operating across multimodal environments and pursuing goals without constant human input, agentic AI introduces both opportunities to advance human rights and risks of opaque, distributed responsibility for harms. This paper argues that the UN Guiding Principles on Business and Human Rights (UNGPs) provide a practical foundation for governing agentic AI. We reinterpret the Protect–Respect–Remedy pillars to map onto agentic capacities: reasoning requires anticipatory due diligence; acting demands safeguards and oversight; and interacting calls for transparency, remedy, and accountability. We propose shared responsibility frameworks, grounded in interoperability standards, ongoing monitoring, and redress mechanisms, as essential governance tools. Ultimately, embedding human rights frameworks into the design and deployment of agentic AI enables both risk mitigation and the active promotion of rights in complex, multi-actor ecosystems.