After several years of intense technological adaptations in the translation industry, there is a need to take stock of their implications, especially as regards the integration of machine translation (MT) in work processes. This study presents the results of a large-scale survey on the use of machine-generated inputs, particularly through translation memories (TMs) and MT systems, in multiple international organizations. It first focuses on the relevance of legal translation in international institutional settings before comparing patterns of use of computer tools for the translation of legal documents as opposed to other texts in these settings, and how such patterns vary across organizations or depending on translators’ profiles. The findings reveal a landscape of widespread “augmented translation”, but with the prevalent use of TMs as suitable tools for verifying the relevance and reliability of the sources of previous translations, while MT suggestions are integrated and post-edited approximately two thirds of the time, or slightly less frequently in the case of legal translation. This points to a more cautious approach to human-machine interaction for the translation of legal texts, which is also reflected in several variations between institutions. From-scratch translation is limited to a minority of cases, while the scores per profile groups based on domain-specific backgrounds and experience levels were strikingly similar across institutions. The factual overview provided serves to debunk some of the myths that have fuelled the hype about MT in recent years.