This paper focuses on the functioning of a surveillance device of resistant bacteria of animal origin, named Resapath. Through a historical and pragmatic analyse, I show how three kinds of actors progressively cooperate with the aim of producing knowledge on antimicrobial resistance risks related to the use of antibiotics in veterinary medicine. But the involvements of microbiologists, epidemiologists and veterinarians in the surveillance device don't rest on the same practices and the same objectives. So Resapth produce three ontologies of resistant bacteria that are partly incommensurable and shape ignorance and uncertainties in the governability of such (non) knowledge objects.