Automated negotiation is especially important when tasks, which require many resources, enter a Grid where resources are scarce. The level of resource scarcity dynamically changes in a Grid and the client's negotiation strategy has to adapt to this dynamism. In addition, we consider the non-transparency of a Grid with respect to a client. That is, a client is only able to observe proposals sent to it by the Grid resource allocator (GRA) but it does not have direct knowledge about availability of Grid resources. In our work, the client's strategy is to estimate the dynamism in a Grid by inferring the criteria influencing the GRA's proposals, and to adapt to this dynamism using fuzzy control rules. These rules define whether the client has to make smaller or larger concessions towards the GRA considering Grid dynamism. The simulation results show that a client who applies our adaptive negotiation strategy can obtain higher utility and significantly reduce the number of failed negotiations comparing to a client who applies the non-adaptive negotiation strategy.