A new study finds UK healthcare workers often assume Google Translate gets medical abbreviations right - a belief the research says is misplaced.
Researchers interviewed 21 healthcare staff across a range of roles and specialties about their use of machine translation, using medical abbreviations as a test case. Abbreviations are already a known source of harm even within a single language, with mix-ups linked to outcomes ranging from minor delays to patient deaths. During semi-structured interviews, participants reviewed French and Spanish examples pulled from authoritative clinical corpora and run through Google Translate, then discussed what they noticed. The responses were analyzed qualitatively to gauge how much staff understood about the tool's limits.
The angle that matters here: UK public healthcare staff already report defaulting to Google Translate to bridge language gaps with patients, often with no institutional sign-off. Layer ambiguous, context-dependent abbreviations on top of ordinary translation uncertainty, and a tool built for travel phrasebooks starts making decisions it was never designed to make.
Google Translate was not built for clinical shorthand, and nothing here suggests it should be trusted with it. If a free consumer app is quietly plugging gaps left by thin interpreter budgets, that is less a story about AI risk and more about a health system asking software to cover for a staffing problem.