As AI is advancing faster than the institutional and social capacities to manage it we have to address its systemic dimension of risks, not only technical or malicious ones.