The spread of smart applications touches the foundations of the criminal law, notably causality, wrongfulness, and legal personhood. First, distributed multi-agent systems form hybrid networks that exhibit emergent behaviours that cannot be attributed to either one of the agents or explained in terms of an aggregation of actions. This makes it hard to determine which action actually caused the harm or the damage that would normally be addressed by the criminal law. Second, it is hard to imagine a smart environment or infrastructure itself becoming the culprit of a criminal charge, but at some point we may have to concede that the self-management that was inscribed in their programs has actually generated a self that should be called to account for its actions. Such a self would require a measure of empathy and a capacity to reflect on the meaning of one’s actions. If artificial agents develop into artificial life forms they would have to share our vulnerability and the ambiguity of our language for us to call them to account in a criminal court.