AI in corporate decision-making: Challenging traditional agency theory and the need for an AI governance framework

In the development history of corporate governance, legal frameworks governing this field have been constructed based on the assumption of humans exercising managerial power and taking responsibility for corporate decisions. With the rapid advancement of Artificial Intelligence (AI), the structure of conventional corporate governance is dramatically changing, for instance AI is widely applied during the recruitment process, credit assessment for risk management and strategic planning. AI is not only a technical supportive tool, but it also is increasingly influencing, shaping and even driving corporate decision-making processes. This shift in company decision-making mechanism raises a new legal question on the modern corporate governance paradigm. If a business decision is critically influenced by AI which causes any damage to the company, its shareholders and relevant stakeholders, who should be charged the responsibility. An argument is provided that AI is deployed for referencing purposes and final decisions still belong to humans’ activities, the link between human judgment and algorithmic influence is becoming increasingly unclear. Particularly, when AI produces inaccurate forecasts, outputs, or misleading recommendations despite accurate input data, it becomes difficult to determine whether responsibility should be charged to the board, the company, or the AI developer. Consequently, conventional mechanisms to address the accountability and managerial oversight of company directors and other managerial positions are facing challenges in the age of AI-driven governance.

Relying on Agency theory, this paper argues that the emergence of the “Algorithmic Agency Problem” is one of ‘game-changer’ factors impacting the foundation of the corporate governance paradigm, which was originally designed to control human behavior rather than algorithmic influence. By applying a comparative legal study on Germany, China, and Vietnam regulatory regime, this paper highlights the different approaches to AI governance, ranging from human oversight models and direct algorithm regulation to governance gaps in developing digital economies. It eventually points out a shift from controlling human behaviour to ‘algorithmic influence’ in the field of corporate governance, and then proposes recommendations for building an efficient AI governance framework for Vietnam.

KEYWORDS: Algorithmic Agency Problem, AI Governance, Corporate Governance, Corporate Decision-making.

This paper was presented at the International Conference titled 'International Trade, Investment, and Financial Law in the Digital Era: A Comparative Perspective between Developed and Developing Nations', held at Foreign Trade University – Ho Chi Minh City Campus (FTU).