This correspondence considers a wireless open-spectrum communication system with N orthogonal narrowband tones and K users who compete to maximize their individual data rates. The correspondence proposes a robust version of the iterative water-filling algorithm (IWFA) that takes into account potential inaccuracies in the noise-plus-interference levels available at the transmitters. It is shown that, in comparison with the conventional version, the robust IWFA (RIWFA) can lead the users to behave in a less greedy fashion and to collaborate unintentionally in order to increase the network sum-rate. In particular, using an appropriately chosen RIWFA parameter, we show that RIWFA yields almost the same sum-rate as price-based IWFA (PIWFA) which requires collaboration of users.