This is the only non-paywalled Englis-language source I found; the facts tally with the wider reporting in German-language news. The event reported took place on 13 May.
Background: In German regional train services the trainset is sometimes split up at an intermediate station - the train from A being split (the automatic couplers disengaged) at station B, the two halves continuing to C and D.
An Iraqi man traveled on such a train and wanted to enquire of a fellow passenger whether the train was to be separated (because in that case he would need to be sure to continue in the correct half). The app he used for Arabic to German translated the idea of the train separating to it exploding (viewed abstractly that makes some sense; an explosion is a kind of separation).
The fellow passenger (understandably IMO) excused herself and raised the alarm. The passenger was arrested without resisting and the train was evacuated and searched for explosives. The passenger had to endure a night in police custody.
Such stories make me question the advisability of using machine translation where I do not read the target language well enough to check the plausibility of the result.
Unfortunately the source text was never reported (a photograph may be in the police files of the passenger’s interview). The passenger was reported as speaking a rare Kurdish dialect so his Arabic input might have been idiosyncratic. From the reporting the police seems to have accepted the translation mistake explanation. The German text as reported second-hand by the witness raising the alarm was “Dieser Zug wird bald explodieren”.
Well that’s pretty bad. Google translate makes that “The train will explode soon.” Seems like that might just be part of it. To make sense it would be followed by something like “Do you know where I stand for train line B” or whatever the options were.
The article doesn’t give the exact quote in either language, but perhaps a native German speaker could give us some idea how a question could have ended up sounding like an announcement. “Does this train get exploded in Hof?” is what I am hearing in my mind’s ear, and that’s odd enough to provoke clarification rather than panic, in ordinary circumstances*. Perhaps the person he asked (via his phone translator) was not familiar with that kind of machine translation, and thought it was a recording, or another person speaking at the other end of a phone connection. (edited to add: could that reported German sentence be either a question or a statement, depending on intonation?)
*and yes, I have to wonder if a different foreigner (me, for example) doing the exact same thing would have gotten the same response, from the report to the overnight police custody.
Do not trust automatic translation! Speaks the professional interpreter (who is biased, of course). Traduttore tradittore (translator - betrayer) is the classic warning, but it applies even more to machine translation.
And if you have to use it and have no way to check the accuracy, do not use google or ChatGPT. Use DeepL. Less bad, but still not perfect.
And yes, I have read about that story in the German press too, just as the OP has told it. It was a text translation and it was frightening to read. The train was searched, the station evacuated, many people lost connections and arrived late. Lots of trouble for a misunderstanding. Now sue the app.
The one linked here is not specific about that. He typed in his language, and his phone “announced” an explosion. Also, he would have had to get quite close to her for her to read it. Perhaps other accounts are more specific on this point. If it was text, that seems even less excuse for panic.
Makes me think of the classic school boy (pre-computer translation) mistake of saying “Ich bin heiß” which IIRC means I am horny (as opposed to “Ich habe heiß”), a cause of endless hours of entertainment as a teenage boy learning German. Though I notice Google Translate will make the same mistake when presented with “I am hot”.
Correction: Mir ist heiß is what you want to say. Ich habe heiß would be the nonsensical “I have hot”.
ETA: And “I am hot” can mean “I am horny”, can it not? If you feed ambiguous sentences into the machine the results will be random.
In English, “I am hot” would mean that I make other people horny. Unless the phrase were further expanded, like “I am hot for so-and-so”, in which case so-and-so makes me horny.
Several years ago I was told a story involving a translation to an Arabic dialect (if I was told more specifically which, I don’t recall) of the word EXIT for an emergency exit sign on a plane.
Apparently the customer was rather upset, because the form of the term used turned out to have a crude meaning in their dialect more like “evacuate yourself” or “shit yourself”.
Those signs got changed out fairly quickly, at no added cost to the customer of course!
I also remember laughing about a decade ago when I tried to use Google translate to help me express a somewhat technical phrase about an airplane from English to French. I don’t remember the phrase but it had the word “fly” in it, which Google helpfully translated as “mouche” (the insect).
I think I just asked a colleague about how to phrase it instead.
This comes from the early days of machine translation, when researchers struggled with how to get translation programs to understand context. Another oft-cited example of this problem is the sentence “Time flies like an arrow”, which has a clear meaning to a human but technically has at least five different possible meanings, all of them equally semantically valid.
Machine translation has been steadily improving over the years, but possibly the most impressive example today is ChatGPT. I can type in the sentence “The spirit is willing but the flesh is weak” which back in the day produced the above-quoted nonsense when translating between English and Russian, and GPT can not only successfully translate it, but explain, on request, what the maxim is supposed to mean.
This article discusses other issues with translation apps
Respond Crisis Translation volunteers say they have seen cases of asylum applications being denied because the translation tool interpreted an “I” in a refugee’s statement as “we”, making it seem as if it was an application for more than one person.