Addressing the Artificial Intelligence Governance Gap in Environmental, Social, and Governance Standards

Abstract

Artificial intelligence (AI) has become essential to corporate decision‐making, yet current environmental, social, and governance (ESG) frameworks offer limited tools for assessing algorithmic responsibility. This paper examines whether AI's distinctive features, namely, lack of transparency, delegated agency, and ongoing adjustment, challenge the organizational reasoning of ESG frameworks. Drawing on organizational theory and science and technology studies (STS), we argue that ESG frameworks, which evolved through incremental adjustment, may prove insufficient for governing algorithmic systems. While integration succeeded for issues such as cybersecurity and climate risk, AI differs because algorithms operate as social and technical systems that distribute responsibility across human and non‐human networks. We propose adding a fourth pillar, Algorithmic Governance, within an extended ESGA framework to address risks that transcend traditional governance categories. It is intended as a conceptual extension of investor‐facing ESG architectures rather than a replacement of existing standards. This pillar highlights fairness, transparency, responsibility, and robustness as core dimensions of corporate responsibility. The paper contributes to organizational theory by conceptually examining conditions under which established governance architectures require structural extension and to technology governance by rethinking responsibility in mixed human‐algorithmic systems. We further discuss how algorithmic risks may vary across environmental, social, and governance domains and outline conceptual approaches for handling heterogeneity, sectoral differences, and data constraints.