We conducted a series of strain-controlled experiments to study the characteristics of a shear zone forming in dense flow of confined dry granular media. The primary objective was to link force fluctuations due to jamming and force network reformation with episodic release of elastic energy as passively monitored by acoustic emission sensors. Under constant deformation rate, the shear stress exhibits a characteristic sawtooth behavior reflecting the strong influence of micromechanical processes on the macroscopic stress-strain behavior. Measured shear stress jumps were highly correlated with low-frequency (< 20 kHz) acoustic emission events. High-frequency (30 kHz–80 kHz) acoustic signals that were measured with different sensors appear to be directly linked to continual grain-scale interactions (e.g., friction, rolling). A conceptual mechanical fiber bundle model (FBM) was used to represent dynamics at the shear zone of large granular assemblies. The model was capable of reproducing the dynamics of stress jumps and associated elastic energy release events. The combination of acoustic emission (AE) measurements and FBM framework offers new insights into the behavior of shear failure and enhances capabilities for resolving grain-scale mechanical processes and for predicting rapid mass movement such as shallow landslides and debris flows.