Trust and deception have been of concern to researchers since the earliest research into multi-agent trading systems (MATS). In an open trading environment, trust can be established by external mechanisms e.g. using secret keys or digital signatures or by internal mechanisms e.g. learning and reasoning from experience. However, in a MATS, where distrust exists among the agents, and deception might be used between agents, how to recognize and remove fraud and deception in MATS becomes a significant issue in order to maintain a trustworthy MATS environment. This paper will propose an architecture for a multi-agent trading system (MATS) and explore how fraud and deception changes the trust required in a multi-agent trading system/environment. This paper will also illustrate several forms of logical reasoning that involve trust and deception in a MATS. The research is of significance in deception recognition and trust sustainability in e-business and e-commerce.