Can AI Interpret in Healthcare? The Boundaries Are Being Drawn

Cross-language communication needs in healthcare are vast and non-negotiable: immigrant patients, international medical tourism, cross-border telemedicine. In 2024, as AI speech translation became more usable, more and more healthcare institutions faced the same question: can AI translation apps be used with patients who don't speak the local language?

Health systems and professional bodies in multiple countries issued guidance this year, and the boundaries are clarifying. Yes: low-risk scenarios like wayfinding, appointment registration and daily inpatient communication, where AI tools meaningfully improve efficiency and ease the shortage of medical interpreters. With caution: history-taking and medication guidance, where AI may assist but key information needs read-back confirmation. No: high-risk scenarios — delivering diagnoses, surgical informed consent, psychiatric assessment — require professional medical interpreters; any mistranslation there can directly endanger patient safety and create legal liability.

The core of this tiered logic is risk matching: AI translation's error rate may be only a few percent, but healthcare's特殊性 is that for whoever those few percent land on, the error is one hundred percent.

For the translation industry, healthcare's boundary-drawing is a template: technology extends the reach of language services, while professional institutions' value lies in holding the moments that "cannot go wrong".