International audienceWe use Abstract Interpretation to automatically prove safety properties of mobile ambients with first order communications. We introduce a non-standard semantics in order to distinguish different recursive instances of agents. This allows us to specify explicitly both the link between agents and the ambient names they have declared, and the link between agents and the ambients they have activated. Then we derive from this non-standard semantics an abstract semantics which focuses on interactions between agents. This abstract semantics describes non uniformly which agents can be launched in which ambients and which ambient names can be communicated to which agents. Such a description is required to prove security proper...