This paper investigates a nonlinear queuing system of type M/M/1/n in which the service rate depends on the current loss probability. A system of Kolmogorov differential equations is derived and numerical experiments are conducted using the classical fourth-order Runge–Kutta method. The results show that increasing the proportionality coefficient between the loss probability and the service rate reduces losses, increases the probability of the idle state, and shortens transient processes. The proposed model reflects adaptive properties of real technical and organizational systems and can be applied to the design of resource management mechanisms with self-regulation.