CO2 geo-sequestration in unminable coal seams is one promising method to mitigate greenhouse gas emissions and would be a means of cleaner production. At the same time, the injected CO2 can enhance coalbed methane recovery and offsets the economic cost for CO2 transportation and injection. However, this technology is still at its primary stage and so far, no large scale of field CO2 injection has been implemented. One of the main difficulties encountered in this process is the permeability reduction induced by CO2 injection and as a result, the injection rate is sharply dropped in the later period. In this paper, detailed laboratory experiments are conducted to investigate the permeability reduction effects during CO2 injections. The dynamic permeability reduction effects are measured with the continuous CO2 injection under different confining pressure conditions. Permeability reduction effects of 13 %���77 % are obtained with respect to the CO2 exposure time in different testings. In low permeability scenario, the permeability reduction effects are severer. Specifically, when the absolute permeability is 1mD, 0.42mD, 0.18mD and 0.03mD, the permeability reduction is 13 %, 36 %, 48 % and 77 %, respectively. Compared with the coal triaxial adsorption tests, it is suggested that the permeability equilibrium is not equal to the adsorption equilibrium. The triaxial adsorption equilibrium process takes much longer time than that of permeability equilibrium process. Based on the testing results, a novel dynamic permeability model is proposed to illustrate the permeability evolutions. The existence of weak internal surface influences the matrix strain and the permeability. This model is also validated by other's experimental results and can be used for explaining coal seam permeability reduction effects of CO2 injection in coal seams.