The current fiscal decentralization system has led to significant differences in regional economic policy uncertainty (economic policy uncertainty, EPU). Due to the limitation of EPU indices being available only at the national level, researchers are unable to examine the impact of regional EPU on domestic capital flows. To address this, a measurement method for regional EPU based on big data and artificial intelligence is proposed. This method empirically investigates the effect and mechanisms of regional EPU disparities on cross-regional capital flows, offering a new explanation for the “Lucas Paradox”, a capital flow anomaly. The study reveals that greater regional EPU disparities in China lead to increased cross-regional capital flows. Analyzing the direction of capital flows shows that the relationship between regional EPU disparities and capital flows is mainly reflected in movements from less developed regions to developed regions. Capital tends to flow from regions with higher EPU to those with lower EPU: enterprises in low-EPU regions are less inclined to actively transfer capital outward, while enterprises in high-EPU regions are more likely to transfer capital outward to mitigate risks. Further analysis indicates an asymmetrical impact of financial development levels and credit levels on the capital-exporting and capital-receiving regions. For capital-exporting regions, local financial development and credit levels weaken the positive relationship between regional EPU disparities and the scale of cross-regional mergers and acquisitions. In contrast, the effect is reversed for capital-receiving regions. Finally, the article offers policy recommendations for the government and enterprises, including reasonably planning pilot areas for various economic policies, maintaining consistency in economic policies, vigorously promoting the development of the financial industry, emphasizing the role of social trust, and improving decision-making mechanisms for cross-regional mergers and acquisitions.