Agent is a complex system that behaves like being alive. It is an emergent property of a complex system that has autocatalytic function. Without autocatalytic function with the ability to perform at least one work cycle there is no life as Stuart Kauffman explained, and without life there is no agent.
What is important consequence of such comprehension of life is that metabolism is not necessary mechanism that supports life. Non-metabolic mechanisms should in theory and in practice produce life as much as metabolism. And since there can be vastly many non-metabolic mechanisms that produce life, it is highly unlikely that humans could find life in Universe by searching for traces of amino-acids, water, earth-like temperature and and other constituencies of metabolic life.
But there is another important consequence coming out of Kauffman conceptualization of life: one can not take work cycle without some kind of an agent. Work cannot be but performed by something. That something is called agent. In theory agents could come in many forms and appearances. Regardless of optionality of their appearance, there is one non-optional feature of any possible agent: responsibility. Responsibility for effects of work cycle on its environment. For this reason we can conclude that ethics applies for all possible agents, for non-metabolic agents as well. Any work produces changes, and any change makes the agent of that change responsible for that change.
Agents are agents even if they are not subjects. Some agents become subjects. Subject is nothing but meme complex assigned to an agent.
Subject can be conscious, but consciousness is only possible emergence on a top of subject’s meme complex.
It follows then that it is not conscious subject, not even any subject that is responsible, but agent only. Ethics should thus be assigned to agents.
Let me explain this result that looks like it is from another reality and not the one that we are part of. If unconscious person hits a person by his car and kills him, most legal systems would not take him responsible or would at least not take him as responsible as if he was conscious in the moment of the accident. If automaton hits and kills a person, automaton without subjectivity, such automaton would not be taken responsible. Now let’s assume that such automaton has a kind of consciousness. Let us assume that unconscious driver becomes conscious a day after an accident. Both the “unconscious” driver and automaton would feel responsible deep inside even if they would be fully discharged by the court and by the public. Socially enacted justice, as important as it is in the process of civilization, has very low if any influence on how agent perceives effects of its actions on his environment. Even Kant’s moral imperative falls short of agent’s responsibility for that responsibility does not happen in memetic, but in personal existential reality.