AI applications raise dozen ethical dilemmas. Shall a self-driving car follow its lawful trajectory and lethally hit a young biker, or shall it stray from its course, hit a tree and kill its old passenger? Similarly, what shall we do when an AI application reveals hidden correlations which denote human biases in society? Those issues necessitate the specification of initial ethical values. In addition, at the instrumental level, a discussion must take place on the methods used to address ethical challenges – self-regulation, formal standardization, statutory obligations – and the stages at which this ought to be done – ex ante by design or ex post by training.