The Dual Impact of Artificial Intelligence on ESG Strategies of Technology Enterprises: Enabling Effects, Governance Risks, and Moderating Mechanisms

Wang Cheng, Bolkvadze Nataliia, Zhang Qian, Deng Yanru

West Ukrainian National University Ternopil

Abstract：

The explosive growth of generative artificial intelligence (AI) has created a core contradiction for technology enterprises' environmental, social, and governance (ESG) strategies. While AI accelerates corporate carbon neutrality and sustainable development, its massive energy consumption offsets long-term emission reduction efforts. This study systematically explores AI's dual impact on corporate ESG performance, transmission channels, moderating conditions, and governance risks. Based on two complementary samples—2,156 firm-year observations of 15 global leading tech enterprises (2020–2026) and 23,844 firm-year observations of Chinese A-share listed tech firms (2011–2022)—we adopt static panel regression, difference-in-differences (DID), and mediating effect tests. Results show: AI application significantly promotes overall ESG performance (β=0.198, p<0.01), with dimensional heterogeneity (environmental > social > governance); High-autonomy AI agents harm governance performance (β=-0.156, p<0.01); Low AI model energy efficiency inhibits environmental performance (β=-0.198, p<0.01); Financing constraint mitigation (28%), green technology innovation (35%), and information disclosure improvement (22%) are key mediating paths, explaining 85% of the total effect; AI exacerbates ESG inequality across industries and regions. We propose a "human-in-command" governance system and eight operable policy recommendations, enriching digital economy and ESG research and providing practical references.

Key Words：

Artificial Intelligence; ESG strategy; dual impact; energy efficiency paradox; agent autonomy; mediation effect; difference-in-differences