Most codes of ethics stipulate that court interpreters should give verbatim renditions, should not have side conversations with the witness or the defendant, and should use the alien-I. However, when we find these maxims flouted by outsourced interpreters working in trials in Barcelona, the observed practices may be considered non-standard and yet constitute an expected and even accepted social practice. Here we attempt to understand why interpreters sometimes abandon all illusion of equivalence, why side conversations occur in certain hearings, and why interpreters sometimes speak in their own voice, becoming direct participants in discursive exchanges. Risk analysis enables us to model ways in which these practices can ensure that cooperation is achieved and time is not wasted. In one case study, side conversations between the defendant and the interpreter serve to inform the defendant of the possible consequences of a plea. Such a practice nevertheless requires that the officers of the court trust interpreters to exceptionally high degrees. In a second case study, disagreements between the judge and the interpreter, technically over issues of translation equivalence, lead to distrust in the interpreter to the point where cooperation becomes impossible. In this instance, a non-standard practice that might be efficient elsewhere leads to communicative failure. It is thus found that non-standard interpreting can be efficient when the participants’ risk-management strategies are aligned and trust is operative; on the other hand, it can also convert trivial differences into high-stakes disputes that throw risk-management strategies out of alignment.