In most works on negotiation dialogues, agents are supposed to be ideally honest. However, there are many situations where such a behaviour cannot always be expected from the agents (e.g. advertising, political negotiation, etc.). The aim of this paper is to reconsider the role of deceitful arguments in argumentation frameworks. We propose a logical tool for representing and handling deceitful arguments in a dialogue between two formal agents having to face their respective knowledge and trying to convince each other. Xlogics, a nonmonotonic extension of classical propositional logics, is used as the background formalism for representing the reasoning of the agents on arguments. Starting from a previous work dedicated to the generation of new arguments, we propose to define the notion of lie as a new kind of possible agent?s answer. Finally we describe the way an agent may trick and how the other agent may detect it.