Despite the rapid global expansion of artificial intelligence (AI), its environmental impacts across socio-economic contexts remain underexplored. This study evaluates the relationship between AI intensity and environmental performance across 56 countries, (2023–2013) using panel econometric methods including MMQR, system GMM, and dynamic threshold modeling. The results reveal that AI activities, especially in high-intensity regimes, are significantly associated with increased CO2 emissions and environmental degradation, confirming the presence of a Digital Rebound Effect. This study identifies a critical threshold beyond which AI's environmental footprint intensifies, particularly in upper-middle-income economies, while lower-middle-income countries experience a technological leapfrogging paradox, where adoption outpaces regulatory capacity. However, stringent environmental policies, a diversified energy mix, and strong digital infrastructure were found to be mitigating these adverse effects. Our findings demonstrate that AI's environmental role is conditional on national policy and productive capabilities. We argue for integrated environmental-digital governance to ensure AI advances sustainability rather than exacerbates ecological risks.
