This paper considers an underlay access strategy for coexisting wireless networks where the secondary system utilizes the primary spectrum to serve its users. We focus on the practical cases where there is uncertainty in the estimation of channel state information (CSI). Here the throughput performance of each system is limited by the interference imposed by the other, resulting in conflicting objectives. We first analyze the fundamental tradeoff between the tolerance interference level at the primary system and the total achievable throughput of the secondary users. We then introduce a beamforming design problem as a multiobjective optimization to minimize the interference imposed on each of the primary users while maximizing the intended signal received at every secondary user, taking into account the CSI uncertainty. We then map the proposed optimization problem to a robust counterpart under the maximum CSI estimation error. The robust counterpart is then transformed into a standard convex semi-definite programming. Simulation results confirm the effectiveness of the proposed scheme against various levels of CSI estimation error. We further show that in the proposed approach, the trade-off in the two systems modelled by Pareto frontier can be engineered by adjusting system parameters. For instance, the simulations show that at the primary system interference thresholds of -10 dBm (-5 dBm) by increasing number of antennas from 4 to 12, the secondary system throughput is increased by 3.3 bits/s/channel-use (5.3 bits/s/channel-use).