{"id":804779,"date":"2026-10-16T19:14:16","date_gmt":"2026-10-16T19:14:16","guid":{"rendered":"https:\/\/www.newsbeep.com\/uk\/804779\/"},"modified":"2026-10-16T19:14:16","modified_gmt":"2026-10-16T19:14:16","slug":"why-a-i-translation-apps-can-fail-doctors-and-patients","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/uk\/804779\/","title":{"rendered":"Why A.I. translation apps can fail doctors and patients."},"content":{"rendered":"

Sign up for the Slatest<\/a> to get the most insightful analysis, criticism, and advice out there, delivered to your inbox daily.<\/p>\n

After one of his patients died, Stephen Hibbs believed A.I. could have saved him\u2014at least, at first. His patient, a young Albanian man, died after a bone marrow transplant in East London. He was discharged with a complex postsurgery medication regimen to keep his immune system in check and prevent infection, one that had to be followed perfectly. But he stopped taking one of the medications too soon and he died before he could explain exactly why he stopped. Hibbs, a blood-cancer specialist, couldn\u2019t stop thinking about it. His patient wasn\u2019t fluent in English. Would discharge instructions in his native Albanian have been easier to follow? Could his death have been avoided? Could others?<\/p>\n

Even when everyone speaks the same language, patients and doctors struggle to communicate about symptoms, treatment options, and risks. Adding language differences raises the odds of things going sideways. Anyone can be vulnerable, including while traveling, or with a cross-lingual medical team. With the rise of generative artificial intelligence tools like ChatGPT and translation apps like Google Translate, healthcare providers have what seems like a cheap and easy way to bridge the gap in their lab coat pockets. But while these technologies promise to make medical communication easier and more affordable, they also introduce new risks of their own. And in a setting where a mistranslated word can mean the difference between an accurate diagnosis and a dangerous mistake, the consequences can be severe\u2014or even deadly.<\/p>\n

In the U.S., people with limited English proficiency bear the most risk. A 2024 research review<\/a> in the journal Healthcare found that those with limited proficiency in English are more likely to experience serious harm from medical errors and receive lower-quality healthcare. That\u2019s one reason federal law requires most providers to make healthcare communication meaningfully available in the language patients understand, via a human interpreter for spoken communication and a translator\u2019s review for written information.<\/p>\n

Yet finding a qualified human isn\u2019t easy. Some 27 million Americans have low English proficiency, yet the entire U.S. translation and interpretation profession numbers just under 74,000, including plenty of people who work in non-healthcare specialties, according to the Bureau of Labor Statistics<\/a>. When President Donald Trump proclaimed English the U.S.\u2019s official language in 2025, these issues worsened. The administration couldn\u2019t repeal healthcare translation laws singlehandedly, Lynette Martins, a professor of health law at DePaul University, told me. But it could deprioritize enforcing those laws, and it changed critical government policies that had previously supported language access. Martins describes it as an \u201cexacerbation of inequality.\u201d<\/p>\n

It wasn\u2019t exactly subtle. The federal government\u2019s new policy is to avoid providing translation services for people with LEP, unless the letter of the law or \u201cmission-critical\u201d practicality demands it. Some of that is anti-woke virtue signaling, but there have also been serious, practical effects. For years, the Department of Justice\u2019s Civil Rights division maintained a resource-rich LEP.gov website, which offered widely used information on how to support people with limited English proficiency, with prominent translations available to 20 other languages commonly spoken in the U.S. But it has been a husk<\/a> for over a year. The single page that remains has a link to Trump\u2019s \u201cEnglish Only\u201d executive order<\/a>. The only other link on the site is to former Attorney General Pam Bondi\u2019s implementation memo<\/a>, which promised new LEP guidance within 180 days. That would have been Jan. 10, 2026. The DOJ didn\u2019t respond to my emailed request for an updated timeline.<\/p>\n

Bondi\u2019s memo also threw shade by implication at the use of human translators, while boosting A.I.: It praised \u201cartificial intelligence and machine translation\u201d as \u201ccost-effective methods for bridging language barriers and reducing inefficiencies with the translation process,\u201d and called for phasing out \u201cunnecessary multilingual offerings.\u201d But many doctors already know what medicine is like without human translators.<\/p>\n

A patient arrived at a Denver emergency room one night terrified that he was having a heart attack. In Spanish, he told William Mundo this was his second ER visit in as many days and handed him the discharge instructions he\u2019d received. Though he was early in his residency, Mundo, a doctor who is bilingual, was familiar with this \u201cbounce back.\u201d Patients with LEP are more likely to make an unplanned return to the hospital after an ER discharge, a 2024 University of California, San Francisco review published in BMJ Quality & Safety<\/a> found.<\/p>\n

He also saw what caused the false alarm: The Spanish translation of how to distinguish expected acid reflux from possible heart attack symptoms was clumsy and confusing. To his eye, the telltale errors of off-the-shelf A.I. translation tools were obvious.<\/p>\n

He understood why his colleagues used them, even though hospitals generally don\u2019t endorse the use of unvetted tech. Whenever his patients spoke a language he didn\u2019t speak, \u201cI did what everybody else would do: copy, paste, throw it onto Google Translate and hope that it was somewhat accurate,\u201d he told me. According to Mundo, hospital culture puts pressure on doctors to \u201cmove the meat.\u201d<\/p>\n

\u201cMove the patients around, keep the hospital moving,\u201d he explained. Translation services, when available, take time most clinicians feel they can\u2019t spare.<\/p>\n

The American Medical Association recently found that 57 percent of doctors<\/a> are using, or plan to use, A.I. translation tools. It\u2019s not because they think it\u2019s ideal, or even OK, says Mundo, who noted he is now senior enough to push back against the time pressure. \u201cA lot of my colleagues tell me, \u2018Well at least [patients] can read the discharge instructions.\u2019 \u201d He can see the justification of \u201ca little bit is better than nothing,\u201d but he\u2019s also seen the harm caused by translations. \u201cIt\u2019s an ongoing ethical dilemma,\u201d he said. \u201cWe try to do the best that we can with what we have.\u201d<\/p>\n

So the medical use of unsanctioned A.I. translation tools is something of an open secret. But it\u2019s hard to measure how dangerous it is in practice, since the field has yet to settle on a standard way to measure A.I. translation safety\u2014even for the A.I. tools that hospitals and health systems have approved for translation use, Dipak Patel, CEO of GLOBO Language Solutions, told me. His company is one of a small group of language-services vendors who have long provided translation services in medical settings, via video, phone, and in person. Several, including GLOBO, now offer A.I. translation. Across the small sector, these services draw on a usually unspecified mix of third-party language models and proprietary technology.<\/p>\n

\u201cIf you ask Hospital A, Hospital B, or any federally qualified health center how they measure the accuracy of interpretation, they\u2019ll all give you the same answer: We let the vendor tell us,\u201d Patel said. With every language-services company using its own accuracy measures, everyone essentially grades their own homework, using their own rules. Without an agreed-upon standard measurement, industrywide accuracy comparisons are impossible.<\/p>\n

And existing A.I. translation metrics aren\u2019t really designed for this task. These often measure text similarity or overlap, but they don\u2019t heavily penalize the tiny errors that should fail in a health setting. It doesn\u2019t help that doctors tend to use language in their own way. For example, when a doctor wants you to stop taking a medication, they\u2019ll often instruct you to \u201chold\u201d it, as a synonym for a temporary pause or delay\u2014like asking someone to hold the elevator. But \u201chold\u201d can have the opposite meaning in a different context: \u201chold your course\u201d means to\u00a0maintain your direction. A mistranslation of that \u201chold\u201d instruction as \u201cadhere to your medication\u201d dangerously reverses the meaning, while being similar enough that it might score decently. (This was an actual A.I. translation error regarding kidney medication\u00a0from English into Russian that researchers found in a 2025 BMJ Quality & Safety study<\/a>. The A.I. translated it correctly in Chinese.)<\/p>\n

Inspired by his experiences, Mundo recently presented a study<\/a> of A.I.-translated patient discharge instructions. When doctors reviewed translations with high quality scores, they also found small but frequent mistakes in medication instructions, as well as in warnings about when to urgently return to the ER.\u00a0 The models also struggled to parse negation correctly, as in \u201ctake with food\u201d versus \u201cdon\u2019t take with food.\u201d<\/p>\n

These problems compound when a model is working in a language that is less represented in A.I. training data. For instance, in a\u00a02025 study of pediatric emergency room scenarios in the journal BMC Health Services Research<\/a>, a model mistranslated a verb from Urdu, so a description of a child\u2019s liquid intake read like the child had consumed alcohol\u2014chilling, when you consider the mistake could open a child protective services case.<\/p>\n

But even in English and Chinese, two of the best-represented languages digitally, changing languages isn\u2019t the same neutral act as changing a phone\u2019s language setting. In a 2025 study in Nature Human Behaviour,<\/a> MIT professor Jackson G. Lu and his colleagues found that OpenAI and China\u2019s Baidu models responded differently to the same prompts depending on whether they were written in English or Chinese. In English, the models tended to favor more individualistic and analytical responses, while in Chinese, they placed greater emphasis on relationships and broader context. In other words, the language of a prompt influenced the cultural assumptions underlying its response. The study wasn\u2019t focused on healthcare specifically, but it\u00a0raises important questions about whether patients and doctors using different languages can reliably receive consistent information about diagnoses, treatment options, and medical risks.<\/p>\n

\u201cI doubt that we could remove every cultural tendency from a model, because language and culture are deeply intertwined,\u201d Lu told me.<\/p>\n

At his East London clinic, Stephen Hibbs\u2014who treated the Albanian patient\u2014found many of his colleagues had similar difficult stories of patients being harmed by language barriers. Translation and interpretation services existed, but not always; take-home materials were invariably in English only.<\/p>\n

This was in the early 2020s, the dawn of the smooth-talking chatbot era. Surely such powerful intelligence could translate discharge instructions into a patient\u2019s preferred language? \u201cGreat,\u201d Hibbs told me, recalling his reaction at the time. \u201cHere\u2019s an off-the-peg solution that we can scale up rapidly and all over the place.\u201d He brought the idea to Federico M. Federici, director of University College London\u2019s Centre for Translation Studies. Looking back, both chuckle at what Hibbs now calls his \u201cfairly naive hope.\u201d<\/p>\n

\u201cIt was not unexpected,\u201d Federici told me. The new LLM tools made translation \u201cseem very easy, very quick.\u201d He suspected they\u2019d be neither\u2014his field had been using machine translation since the 1950s. But both wanted to learn whether A.I. translation could be trusted in high-stakes medicine. They launched a study to find out.<\/p>\n

Hibbs and Federici published their findings in the journal Supportive Care in Cancer<\/a> in March 2026. For the study, a simple booklet called Easy Read: How is Myeloma Treated?<\/a> was translated from English into Bengali twice\u2014once by machine, and once by humans. For the version translated by A.I., a research team member who did not speak Bengali put it through Google Translate. For the human version, the best professional translation practices were applied: One professional translator\u2019s work was checked by another, and a third proofread for readability.<\/p>\n

When both were scored for accuracy by a different group of assessors who didn\u2019t know which translation was which, neither proved to be error-free. Out of a possible 100 points on a scale used by the Chartered Institute of Linguists, the human version scored a 73, losing points mainly on points of language and clarity. As for the A.I. translation, most of the booklet was translated accurately, the assessors found. But its score was still a lousy 51.<\/p>\n

Although there weren\u2019t many of them numerically, the errors that the A.I. did make were confusing and bewildering enough to be disqualifying. For instance, the line \u201cThe medicine will help you to: Kill as many myeloma cells as possible\u201d was mistranslated as instructing the patient to somehow kill as many myeloma cells as possible. There is a specific word for medical \u201cside effects\u201d in Bengali, but the A.I. translated it as \u201charmful side.\u201d Other errors were pure gibberish. \u201cWe can put you in touch with another myeloma patient to talk to\u201d became \u201cWe can contact other patients for us to talk to you.\u201d After reading the weird A.I. translation, a Bengali speaker said it would make them question whether they could trust the clinic\u2014or the doctor who handed it to them\u2014at all.<\/p>\n

But here\u2019s the thing: When they tested comprehension, the group that read the far superior human translation understood just as little as those who read the A.I. version. \u201cAlmost no one understood the key points,\u201d Hibbs said.<\/p>\n

<\/p>\n

\"\"<\/p>\n

\n Steffi Cao
\n Tech Bros Have Found a New Creepy Way to Profit Off Your Health Anxiety
\n Read More\n <\/p>\n

<\/a><\/p>\n

The final component of the study is where that changed. All participants had a live conversation with a clinician and a professional interpreter, simulating the informational discussion of the risks and benefits which typically precedes myeloma treatment. Part of this conversation includes walking through the patient consent form.\u00a0One group reviewed the standard\u00a0English-only consent form, while the other had a professionally translated \u201cinterlinear\u201d form, which placed Bengali and English text near each other. On the test that followed, comprehension improved for both groups, but it was highest for the group whose conversation included the bilingual form.<\/p>\n

In other words, more human time and attention boosted understanding.<\/p>\n


\n This Content is Available for Slate Plus members only<\/p>\n

Doctors Are Using a Shortcut for an Age-Old Problem. It Can Be Pretty Risky.
\n <\/a><\/p>\n


\n This Content is Available for Slate Plus members only<\/p>\n

How Did Hilary Swank Get Pregnant at 52? It\u2019s Not Hard to Understand, if You Know a Little About Biology.
\n <\/a><\/p>\n

Still, the takeaway isn\u2019t \u201cA.I. translation bad, human translation good.\u201d Human translators rely on A.I. tools; they have to, just to meet even a fraction of demand, Federici said. And in the LLM era, human involvement remains a necessary failsafe\u2014not just as an error-catcher, but to ensure meaning survives. \u201cIt\u2019s the unsupervised use of this technology, scaling it in the way that Stephen [Hibbs] was initially describing, that is fundamentally dangerous,\u201d Federici said.<\/p>\n

Hibbs agrees. He started with the hope that A.I. translation could save his patients\u2019 lives. Now that he\u2019s seen the kinds of errors it makes, that hope is gone. Without human review, \u201cI would not use an A.I. tool, even if there was no other alternative,\u201d he said. Unless accuracy improves significantly, he thinks handing a patient a translation that might be wrong is ethically worse than giving them no written information at all\u2014as risky as sending a medical student to perform an operation solo. \u201cIn almost every case, it\u2019d be better to avoid the surgery than to expose the patient to the risk of a botched operation.\u201d<\/p>\n

<\/p>\n

Copy Link
\n <\/a><\/p>\n

<\/p>\n

Gift Story
\n <\/a><\/p>\n

You have remaining gift available this month. This gift link is valid through .<\/p>\n

There was an error with your gift link. Please try again.<\/p>\n

Copy link\n <\/p>\n

Link copied!\n <\/p>\n

Only Slate Plus members can gift Slate stories. Become a member to share 10 free articles a month.Become a Plus Member<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"Sign up for the Slatest to get the most insightful analysis, criticism, and advice out there, delivered to…\n","protected":false},"author":2,"featured_media":804780,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[43],"tags":[733,102,3247,2960,9021,3100,56,54,55],"class_list":["post-804779","post","type-post","status-publish","format-standard","has-post-thumbnail","category-healthcare","tag-artificial-intelligence","tag-health","tag-health-care","tag-healthcare","tag-language","tag-medicine","tag-uk","tag-united-kingdom","tag-unitedkingdom"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/posts\/804779","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/comments?post=804779"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/posts\/804779\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/media\/804780"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/media?parent=804779"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/categories?post=804779"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/tags?post=804779"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}