In this work we address the safe adaptive control problem for autonomous vehicles in the highway on-ramp merging scenario. We argue that by developing a Control Barrier Function (CBF)-based method, autonomous vehicles are able to perform adaptive interactions with human drivers with safety guarantees under uncertainty. We propose a novel extension of traditional CBF to a probabilistic setting for stochastic system dynamics with provable chance-constrained safety and provide a theoretical analysis to discuss the solution feasibility guarantee and design factors reflecting different vehicle behaviors. This allows for adaptation to different driving strategies with a formally provable feasibility guarantee for the ego vehicle’s safe controller. The results demonstrate the enhanced safety and adaptability of our proposed approach.