Cu(In,Ga)Se2 (CIGS) absorbing materials are widely used for thin film solar cells. Recently, it has been found that the addition of silver further improves the efficiency. Increasing the Ag content leads to a decrease in carrier-density, an increase in grain-size, and a flattened [Ga]/([Ga] + [In]) (GGI) profile. The thermodynamic and kinetic properties of Ag in CIGS, however, are only partly understood. In this contribution we employ a machine-learning interatomic potential based on the atomic-cluster-expansion method for (Ag,Cu)GaSe2 to investigate the thermodynamics and kinetics of silver and copper. Specifically, we consider the diffusion properties of prevalent defect species, energy barriers for defect migration and enthalpy of mixing for different compositions of (Ag,Cu)GaSe2. We compare these results to a cluster-expansion model fitted to density functional theory calculations of the mixing enthalpy. Employing the machine learning potential allows us to explore system sizes and time scales typically inaccesible to first-principles calculations.