One of the fastest growing applications of AI research is the implementation of computer programs commonly referred to as ‘agents’. This type of software is distinguished from more traditional programs by a high degree of autonomy in decision making and action, the ability to ‘learn’ from experience and to adapt their behaviour accordingly, and often a highly personified interface. Many are specifically designed to process complex information, make decisions and initiate actions in ‘mission critical’ areas of human endeavour including health, scientific research, government, business, defence, the law and increasingly in education. While in some cases we are aware of our interactions with these electronic entities, in many contexts their activity takes place ‘behind the scenes’, at a level not apparent to the user. Implicit in our conception of an agent both in the physical world and in cyberspace is the notion of delegation. Important aspects of this concept are our understandings of features of human interaction such as trust, responsibility, privacy and our capacity to judge competence and intention. Consideration of these issues in relation to the activities of software agents could lead to the formulation of a broadly based code of ‘agent ethics’. This could help in regulating some aspects of agent behaviour and act as a foundation upon which common expectations on the part of users might be formulated.