A biped humanoid robot is prone to fall when walking or operating in a complex environment, and forward fall is one of the most common falling cases. This study focuses on the forward fall issue and presents an"Arm Flexible Landing Strategy" for safe falling. First, the forward falling motion is analyzed from an energy variation perspective of the robot system, and a method to choose best-fit landing attitude of the arm is presented. Then, a exible landing controller is implemented in the arm to reduce the impact force to the robot, thereby further increasing protection. The presented algorithm is very easy to implement and does not require any additional physical elements to the robot. Finally, a series of simulations are made on a humanoid robot to validate the effectiveness of the presented methods on a safe landing.